[gentoo-user] BIG reiserfs problem

[gentoo-user] BIG reiserfs problem

[CapSel]

It's now more than five times when reiserfs has "sucked my data into
/dev/null". At the begining I thout that was a hardware problem -
disk, ram... but now I am almost 100% sure that reiserfs IS NOT stable
file system. It doesn't matter if I have gentoo-sources or
hardened-sources, if I compile for my arch or for i386... heavy load
or just one rsync process, gentoo or slackware (I thought that I gave
"bad" CFLAGS, USE...).
...it lacked support for SEcurity labels some time ago...
But it is still fastest fs.
Am I the only one who have this problem?
So my question is - how can I help to eliminate this bug(s)?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Uwe Thiem]

On 28 October 2006 12:16, CapSel wrote:
> It's now more than five times when reiserfs has "sucked my data into
> /dev/null". At the begining I thout that was a hardware problem -
> disk, ram... but now I am almost 100% sure that reiserfs IS NOT stable
> file system. It doesn't matter if I have gentoo-sources or
> hardened-sources, if I compile for my arch or for i386... heavy load
> or just one rsync process, gentoo or slackware (I thought that I gave
> "bad" CFLAGS, USE...).
> ...it lacked support for SEcurity labels some time ago...
> But it is still fastest fs.
> Am I the only one who have this problem?

I have used ReiserFS for years without any problems. On gentoo systems as well 
as SuSE.

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
http://www.SysEx.com.na
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Dale]

b.n. wrote:
> CapSel ha scritto:
>> It's now more than five times when reiserfs has "sucked my data into
>> /dev/null". At the begining I thout that was a hardware problem -
>> disk, ram... but now I am almost 100% sure that reiserfs IS NOT stable
>> file system. It doesn't matter if I have gentoo-sources or
>> hardened-sources, if I compile for my arch or for i386... heavy load
>> or just one rsync process, gentoo or slackware (I thought that I gave
>> "bad" CFLAGS, USE...).
>> ...it lacked support for SEcurity labels some time ago...
>> But it is still fastest fs.
>> Am I the only one who have this problem?
>> So my question is - how can I help to eliminate this bug(s)?
>
> It's two years I'm using ReiserFS and I still had no problem *fingers
> crossed*.
>
> You are scaring me, anyway, :) and yours is not the first horror story
> I have heard about ReiserFS.
>
> I've heard good things about XFS, probably when I'll have to format
> I'll give it a try.
>
> m.

If you use XFS, make sure you have good power.  XFS does not like power
failures at all.  I have had to reinstall on a second rig because of
this very problem.  If you have a UPS, that may be OK. 

I have used Reiserfs since I started using Linux four years ago and I
have not had any trouble at all.  I would make sure it is not something
bad on the drive.  I'm even looking forward to Reiserfs4 myself.

Also, exactly what is the failure you have?  Have you googled to see if
others have the same issue and maybe a fix?

Dale

:-)  :-)
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: BIG reiserfs problem

[Alexander Skwar]

· Dale <dalek@exceedtech.net>:

> I'm even looking forward to Reiserfs4 myself.

Well, actually I wouldn't hold my breath for Reiserfs anymore,
now that Hans Reiser is in jail. I really don't think that reiserfs
will have a good future.

Alexander Skwar
-- 
Elwood:  What kind of music do you get here ma'am?
Barmaid: Why, we get both kinds of music, Country and Western.


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Uwe Thiem]

On 28 October 2006 13:14, Alexander Skwar wrote:
> · Dale <dalek@exceedtech.net>:
> > I'm even looking forward to Reiserfs4 myself.
>
> Well, actually I wouldn't hold my breath for Reiserfs anymore,
> now that Hans Reiser is in jail. I really don't think that reiserfs
> will have a good future.

He isn't in jail but in detention. No court has convicted him yet.

Still, ReiserFS4 seems unlikely to happen. :-(

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
http://www.SysEx.com.na
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Mark Kirkwood]

CapSel wrote:
> It's now more than five times when reiserfs has "sucked my data into
> /dev/null". At the begining I thout that was a hardware problem -
> disk, ram... but now I am almost 100% sure that reiserfs IS NOT stable
> file system. It doesn't matter if I have gentoo-sources or
> hardened-sources, if I compile for my arch or for i386... heavy load
> or just one rsync process, gentoo or slackware (I thought that I gave
> "bad" CFLAGS, USE...).
> ...it lacked support for SEcurity labels some time ago...
> But it is still fastest fs.
> Am I the only one who have this problem?
> So my question is - how can I help to eliminate this bug(s)?


I've used Linux for many years - the *only* times I have ever lost data 
has been due to reiserfs file corruption.... and both times I had / and 
/usr mounted as reiserfs, so I'd recommend avoiding it on these two 
filesystems anyway!

I'd recommend changing to ext3 or xfs, as I've found both to be solid (I 
prefer xfs but that just my personal opinion).

Cheers

Mark
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Dale]

[-- Attachment #1: Type: text/plain, Size: 621 bytes --]

Alexander Skwar wrote:
> · Dale <dalek@exceedtech.net>:
>
>   
>> I'm even looking forward to Reiserfs4 myself.
>>     
>
> Well, actually I wouldn't hold my breath for Reiserfs anymore,
> now that Hans Reiser is in jail. I really don't think that reiserfs
> will have a good future.
>
> Alexander Skwar
>   

According to the homepage, it is already released, not finished but
released.

http://www.namesys.com/v4/v4.html

Surely someone will take it over.  It looks promising to me.  Yup, I saw
where he was arrested.  Doesn't mean he did it though.  I'll wait on the
facts. 

Dale

:-)  :-)

[-- Attachment #2: Type: text/html, Size: 1234 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Mark Kirkwood]

Dale wrote:

> 
> If you use XFS, make sure you have good power.  XFS does not like power
> failures at all.  I have had to reinstall on a second rig because of
> this very problem.  If you have a UPS, that may be OK. 
> 

Interesting - I'm running an xfs system that has been through several 
power failures without problems - could have more to do with the 
specifics of ones harddrives than the type of file system (e.g. ATA/IDE 
write caching being the most common reason for power loss corruption, 
and some types seem more susceptible than others).

Cheers

Mark
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 14:42, b.n. wrote:

> You are scaring me, anyway, :) and yours is not the first horror story I
> have heard about ReiserFS.
>
> I've heard good things about XFS, probably when I'll have to format I'll
> give it a try.

than you did not hear of the '2.6.17 kernels XFS eats data' stories. If you 
have xfs and boot any 2.6.17.X with X<5 kernels, XFS will be corrupted.


Most reiserfs horror stories are from early 2.4 kernels anyway.

And if someone has lots of reiser corruption, there are really only two 
options left:

something is wrong with his hardware
he optimized his system to death.

He did not even tell, how he ruled out hardware problems.

I have seen data corruption because of flaky cables, bad PSU, bad harddisks 
(IBM deathstar), bad ram....
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 13:31, Mark Kirkwood wrote:

>
> I'd recommend changing to ext3 or xfs, as I've found both to be solid (I
> prefer xfs but that just my personal opinion).

if you use XFS don't use 2.6.17 kernels.

if you use ext3 don't use 2.6.18 kernels.

...
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[CapSel]

[-- Attachment #1: Type: text/plain, Size: 404 bytes --]

So maybe you know any file system with journaling that will ensure
consistency of *data and metadata* but not "slow" and does not have limit on
amount of subdirectorys (I need almost 100000 subdirs in one dir). I know
that ext3 (resiserfs too) have data=journal (man mount) but it has limit on
amount of subdirs (32000). If it is possible how to enable somthing similar
to "data=journal" in jfs and xfs?

[-- Attachment #2: Type: text/html, Size: 429 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Novensiles divi Flamen]

[-- Attachment #1: Type: text/plain, Size: 360 bytes --]

On Saturday 28 October 2006 18:40, Hemmann, Volker Armin wrote:
>
> if you use ext3 don't use 2.6.18 kernels.
>
Is there a specific problem with ext3 and 2.6.18 kernels? I haven't heard 
anything yet, but if there is a prob it's likely to bite me very soon... 
references?

- Noven
-- 
>-- Novensiles divi Flamen --<
>---- Miles Militis Fons ----<

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[b.n.]

CapSel ha scritto:
> It's now more than five times when reiserfs has "sucked my data into
> /dev/null". At the begining I thout that was a hardware problem -
> disk, ram... but now I am almost 100% sure that reiserfs IS NOT stable
> file system. It doesn't matter if I have gentoo-sources or
> hardened-sources, if I compile for my arch or for i386... heavy load
> or just one rsync process, gentoo or slackware (I thought that I gave
> "bad" CFLAGS, USE...).
> ...it lacked support for SEcurity labels some time ago...
> But it is still fastest fs.
> Am I the only one who have this problem?
> So my question is - how can I help to eliminate this bug(s)?

It's two years I'm using ReiserFS and I still had no problem *fingers 
crossed*.

You are scaring me, anyway, :) and yours is not the first horror story I 
have heard about ReiserFS.

I've heard good things about XFS, probably when I'll have to format I'll 
give it a try.

m.
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: Re: BIG reiserfs problem

[Alexander Skwar]

· Uwe Thiem <uwix@iway.na>:

> On 28 October 2006 13:14, Alexander Skwar wrote:
>> · Dale <dalek@exceedtech.net>:
>> > I'm even looking forward to Reiserfs4 myself.
>>
>> Well, actually I wouldn't hold my breath for Reiserfs anymore,
>> now that Hans Reiser is in jail. I really don't think that reiserfs
>> will have a good future.
> 
> He isn't in jail but in detention. No court has convicted him yet.

Sorry about the confusion. You're right, I didn't know the english
word for that.

Alexander Skwar
-- 
<xinkeT> "Lord grant me the serenity to accept the things I cannot
         change, the courage to change the things I can, and the wisdom
         to hide the bodies of the people I had to kill because they
         pissed me off."


-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: BIG reiserfs problem

[Alexander Skwar]

· Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>:

> if you use ext3 don't use 2.6.18 kernels.

Why?

Alexander Skwar
-- 
A student who changes the course of history is probably taking an exam.


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 16:41, b.n. wrote:
> Dale ha scritto:
> > If you use XFS, make sure you have good power.  XFS does not like power
> > failures at all.  I have had to reinstall on a second rig because of
> > this very problem.  If you have a UPS, that may be OK.
>
> Thanks a lot for the advice. Power outages do happen and I don't have an
> UPS. Why does it happen? Isn't XFS journaled?
>

a) journaling does not prevent data loss. It prevents fs corruption

b) XFS does very aggressive caching. Do you have a power outage, you might be 
lucky, or you might be screwed.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 14:16, Novensiles divi Flamen wrote:
> On Saturday 28 October 2006 18:40, Hemmann, Volker Armin wrote:
> > if you use ext3 don't use 2.6.18 kernels.
>
> Is there a specific problem with ext3 and 2.6.18 kernels? I haven't heard
> anything yet, but if there is a prob it's likely to bite me very soon...
> references?
>
> - Noven

http://marc.theaimsgroup.com/?l=linux-kernel&m=116161490922657&w=2

http://marc.theaimsgroup.com/?l=linux-kernel&m=116184293516235&w=2
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Dale]

b.n. wrote:
> Dale ha scritto:
>
>> If you use XFS, make sure you have good power.  XFS does not like power
>> failures at all.  I have had to reinstall on a second rig because of
>> this very problem.  If you have a UPS, that may be OK. 
>
> Thanks a lot for the advice. Power outages do happen and I don't have
> an UPS. Why does it happen? Isn't XFS journaled?
>
> m.

I had the exact same thing happen twice.  It ran every other file system
just fine, but XFS with one power failure and it was unbootable from
then on.

I don't know why I just know that I used Reiserfs after that and it
worked fine, even after power failures, lots of them too.

That said, nothing is perfect.  A UPS is a good idea even if it can only
last long enough for a proper shutdown.

Dale

:-)  :-)
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 15:03, Alexander Skwar wrote:
> · Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>:
> > if you use ext3 don't use 2.6.18 kernels.
>
> Why?
>
http://marc.theaimsgroup.com/?l=linux-kernel&r=1&b=200610&w=2

go there, search for ext3

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Norberto Bensa]

[-- Attachment #1: Type: text/plain, Size: 288 bytes --]

CapSel wrote:
> So my question is - how can I help to eliminate this bug(s)?

Can you check your RAM please? Reiserfs (3.x that is) is very stable. I'm 
using it for five years now. No data loss or corruption.

And DON'T use XFS if you can't afford an UPS.


Regards,
Norberto

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[fire-eyes]

Norberto Bensa wrote:
> CapSel wrote:
>> So my question is - how can I help to eliminate this bug(s)?
> 
> Can you check your RAM please? Reiserfs (3.x that is) is very stable. I'm 
> using it for five years now. No data loss or corruption.
> 
> And DON'T use XFS if you can't afford an UPS.

Strongly agree on all points there.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[b.n.]

Dale ha scritto:

> If you use XFS, make sure you have good power.  XFS does not like power
> failures at all.  I have had to reinstall on a second rig because of
> this very problem.  If you have a UPS, that may be OK. 

Thanks a lot for the advice. Power outages do happen and I don't have an 
UPS. Why does it happen? Isn't XFS journaled?

m.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Norberto Bensa]

[-- Attachment #1: Type: text/plain, Size: 245 bytes --]

b.n. wrote:
> Yes, but it costs money :)

Not that much really if you think how much it will save :)

You don't need a keep-my-box-up-30-days UPS. A 15 minutes UPS will do just 
fine and they are very cheap nowdays.

Regards,
Norberto

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[fire-eyes]

Norberto Bensa wrote:
> b.n. wrote:
>> Yes, but it costs money :)
> 
> Not that much really if you think how much it will save :)
> 
> You don't need a keep-my-box-up-30-days UPS. A 15 minutes UPS will do just 
> fine and they are very cheap nowdays.

Another nice note is that APC and probably others can be connected via
USB, and configured to cleanly halt your system if its batteries run
out. Always better than a hard dump.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Statux]

[-- Attachment #1: Type: text/plain, Size: 740 bytes --]


> This becomes a philosophical question but, shouldn't computers (hw and 
> sw) be thought to resist hard power outages as much as possible?
> If UPS systems are the only feasible way to obtain it, why no one 
> thought to somehow include them inside desktop systems?

I'm sure that they're out there but not in most applications. It's all
about smaller and lighter now adays with electronics, but you could
probably find a PSU and a UPS combined somewhere; not sure how big it
would be or if it would fit within the ATX formfactor for desktop
systems, though. Now, mounting something off the back of the case is
another story...

I'm getting too creative right now so I'll just stop :)

-- 
Statux <statux@optonline.net>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[b.n.]

Hemmann, Volker Armin ha scritto:
> On Saturday 28 October 2006 16:41, b.n. wrote:
>> Dale ha scritto:
>>> If you use XFS, make sure you have good power.  XFS does not like power
>>> failures at all.  I have had to reinstall on a second rig because of
>>> this very problem.  If you have a UPS, that may be OK.
>> Thanks a lot for the advice. Power outages do happen and I don't have an
>> UPS. Why does it happen? Isn't XFS journaled?
>>
> 
> a) journaling does not prevent data loss. It prevents fs corruption

Well, but having to reinstall a system usually means you've badly 
corrupted your FS. I'm a bit naive maybe -which realistic situation can 
lead to a screwed system by simple data loss?

> b) XFS does very aggressive caching. Do you have a power outage, you might be 
> lucky, or you might be screwed.

Yes, I read the previous thread, and I'm quite perplexed. A cache of 
*megabytes* kept for *hours*? Mmm. I'm not sure it's what I want and/or 
need.

What about JFS?

m.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[b.n.]

> I don't know why I just know that I used Reiserfs after that and it
> worked fine, even after power failures, lots of them too.

Yes, that's one of the reasons I like it.

> That said, nothing is perfect.  A UPS is a good idea even if it can only
> last long enough for a proper shutdown.

Yes, but it costs money :)
I run a fairly simple desktop system, so if I get a power outage rarely 
it's a dramatic thing. Unless I get an unbootable system, in which case 
I can get very angry...

m.

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 18:38, b.n. wrote:

> What about JFS?

it is known to be pretty robust and you will have a hard time to find 
any 'horror stories' - but one reason for the lack of horror stories: there 
aren't many users.
And it is very slow.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[CapSel]

[-- Attachment #1: Type: text/plain, Size: 2235 bytes --]

I wrote five because I started to count 5 times ago. My RAM is in good
condition :) Problem with reiserfs (reiser4 is used across the net to
specify version 4.X IMHO) occurs on many machines - servers, backup servers
(rsync once a day), desktop computers (3 or 4), laptop (only one).
  Every one of them was tested and works with ext3 (but it has limit of
subdirs), every one of them works fine under UFS2 (freebsd, limit of dirs,
too). Some of those servers work hard with mysql,php,apache (more than 2000
queries per second) and errors occurs randomly... 1 per week.
 Backup servers doesn't do very much - only one rsync process at a time and
this is enough to break reiserfs.

To resume:
1. reiserfs breaks down on power failures even with option data=journal or
sync
2. reiserfs breaks down on every machine that I have - from pentium-mmx to
Core2Duo, about 10 machines
3. reiserfs breaks down on heavy load (/proc/loadavg=35), and low load(=0.1),
*it only takes a time*
4. reiserfs broke down on all partitions that was using reiserfs
5. reiserfs breaks down on 2.4.21+ and some older
6. reiserfs breaks down on 2.6.17,  ..18 and older that I can't remember
7. ram, disks, cables was tested, changed, and problem stays
8. ...sata, ata/pata
9. maybe those are just accidents but they occur too often to call them just
accidents

Conclusions:
1. UPS will not help you with reiserfs
2. Do not use reiserfs on production systems
3. "quite stable" means "works for them, not for me".
4. journal only slows things down and does not preserve any consistency

My questions:
1. Is there the file system that preserves data & metadata like UFS2 ????
"solid as rock!"
2. What options do I need to enable (mount/mkfs) for XFS or JFS to enable
FULL journaling?
3. How to bypass limit od 32000 dirs on ext3?
4. Is there an option in kernel and how to enable it to sync on Oops and
reboot?

On 10/28/06, Norberto Bensa <nbensa@gmx.net> wrote:
>
> CapSel wrote:
> > So my question is - how can I help to eliminate this bug(s)?
>
> Can you check your RAM please? Reiserfs (3.x that is) is very stable. I'm
> using it for five years now. No data loss or corruption.
>
> And DON'T use XFS if you can't afford an UPS.
>
>
> Regards,
> Norberto
>
>
>

[-- Attachment #2: Type: text/html, Size: 2657 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 19:16, CapSel wrote:
> I wrote five because I started to count 5 times ago. My RAM is in good
> condition :) Problem with reiserfs (reiser4 is used across the net to
> specify version 4.X IMHO) 

so you are using reiser4?

You know that is not even in a stable kernel and still regarded as testing?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[CapSel]

[-- Attachment #1: Type: text/plain, Size: 572 bytes --]

I use reiserfs (version 3.x, if correctly recall 3.6), haven't even touched
reiser4(resierfs 4.X).

On 10/28/06, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>
wrote:
>
> On Saturday 28 October 2006 19:16, CapSel wrote:
> > I wrote five because I started to count 5 times ago. My RAM is in good
> > condition :) Problem with reiserfs (reiser4 is used across the net to
> > specify version 4.X IMHO)
>
> so you are using reiser4?
>
> You know that is not even in a stable kernel and still regarded as
> testing?
> --
> gentoo-user@gentoo.org mailing list
>
>

[-- Attachment #2: Type: text/html, Size: 928 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Dale]

[-- Attachment #1: Type: text/plain, Size: 1043 bytes --]

Statux wrote:
>> This becomes a philosophical question but, shouldn't computers (hw and 
>> sw) be thought to resist hard power outages as much as possible?
>> If UPS systems are the only feasible way to obtain it, why no one 
>> thought to somehow include them inside desktop systems?
>>     
>
> I'm sure that they're out there but not in most applications. It's all
> about smaller and lighter now adays with electronics, but you could
> probably find a PSU and a UPS combined somewhere; not sure how big it
> would be or if it would fit within the ATX formfactor for desktop
> systems, though. Now, mounting something off the back of the case is
> another story...
>
> I'm getting too creative right now so I'll just stop :)
>
>   

I'm thinking you could make the capacitors on the high voltage side MUCH
larger so it would last longer.  It would still have to be a fast
shutdown though.  Something like shutdown -h -t -5 minutes ago.  LOL 
The only thing about that is charging them when you first turn on the
computer. 

Dale

:-)  :-)

[-- Attachment #2: Type: text/html, Size: 1469 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[b.n.]

fire-eyes ha scritto:
> Norberto Bensa wrote:
>> b.n. wrote:
>>> Yes, but it costs money :)
>> Not that much really if you think how much it will save :)
>>
>> You don't need a keep-my-box-up-30-days UPS. A 15 minutes UPS will do just 
>> fine and they are very cheap nowdays.

Last time I checked they were in the 40-50 Euro range. However you are 
right, it probably can save my day.

> Another nice note is that APC and probably others can be connected via
> USB, and configured to cleanly halt your system if its batteries run
> out. Always better than a hard dump.

This becomes a philosophical question but, shouldn't computers (hw and 
sw) be thought to resist hard power outages as much as possible?
If UPS systems are the only feasible way to obtain it, why no one 
thought to somehow include them inside desktop systems?

Just to know :)

m.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 250 bytes --]

On Sat, 28 Oct 2006 11:21:25 -0300, Norberto Bensa wrote:

> And DON'T use XFS if you can't afford an UPS.

Unless you're using a laptop.


-- 
Neil Bothwick

Approx. 1 in 36000 people will break a leg within 3 weeks of reading this
post

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Joe Menola]

On Saturday 28 October 2006 2:02 pm, Neil Bothwick wrote:
> > And DON'T use XFS if you can't afford an UPS.
>
> Unless you're using a laptop.

Solar UPS?

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 269 bytes --]

On Sat, 28 Oct 2006 14:14:06 -0500, Joe Menola wrote:

> > > And DON'T use XFS if you can't afford an UPS.  
> >
> > Unless you're using a laptop.  
> 
> Solar UPS?

Battery!


-- 
Neil Bothwick

"Bother," said Pooh, as someone flamed him for no reason.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Alan McKinnon]

On Saturday 28 October 2006 16:41, b.n. wrote:
> Dale ha scritto:
> > If you use XFS, make sure you have good power.  XFS does not like
> > power failures at all.  I have had to reinstall on a second rig
> > because of this very problem.  If you have a UPS, that may be OK.
>
> Thanks a lot for the advice. Power outages do happen and I don't have
> an UPS. Why does it happen? Isn't XFS journaled?

Yes it is journaled but it also allows data to be very aggressively 
cached. Make that VERY aggressively cached. With the result that data 
can be held in a huge cache somewhere and the kernel can be convinced 
it has been written to disk.

Consider XFS's pedigree - SGI wrote it for their graphics machines. 
These were big monsters backed up with high grade UPSs and such - the 
logic was that if you spend a brazillion bucks on hardware, a mega UPS 
is part of the deal, along with the wages to pay the army of admins you 
also need.

And, when doing video rendering, it turns out that it's easier to simply 
re-render a frame when the filesystems does something odd with the data 
rather than go to the effort of writing an FS that is 100% reliable. So 
SGI sacrificed something that doesn't actually matter for their use 
case to gain a significant performace increase (which does matter a 
great deal)

alan
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: BIG reiserfs problem

[Alexander Skwar]

· Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>:

[ JFS ]

> And it is very slow.

Based on what evidence/test?

According to http://linuxgazette.net/122/TWDT.html#piszcz it's quite
fast.

Alexander Skwar
-- 
There is nothing new except what has been forgotten.
                -- Marie Antoinette


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Chris Walters]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Alexander Skwar wrote:
> · Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>:
> 
> [ JFS ]
> 
>> And it is very slow.
> 
> Based on what evidence/test?
> 
> According to http://linuxgazette.net/122/TWDT.html#piszcz it's quite
> fast.
> 
> Alexander Skwar

Allow me to add that from my own tests (e.g. benchmarks) and personal
experience with JFS, it is quite fast.

Chris Walters
-----BEGIN PGP SIGNATURE-----

iD8DBQFFQ9FcUx1jS/ORyCsRCOAxAJ9G7FZF89EJOiYsUqhVeKcr6o13ngCgjlFo
sWTOuK4F4BQakNuW9IsMnlA=
=CpJs
-----END PGP SIGNATURE-----
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Jerry McBride]

On Saturday 28 October 2006 17:47, Alexander Skwar wrote:
> · Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de>:
>
> [ JFS ]
>
> > And it is very slow.
>
> Based on what evidence/test?
>
> According to http://linuxgazette.net/122/TWDT.html#piszcz it's quite
> fast.
>


Nice graphs... looking them over make me quite happy I use EXT3 as my 
filesystem of choice....


Jerry

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Saturday 28 October 2006 19:33, CapSel wrote:
> I use reiserfs (version 3.x, if correctly recall 3.6), haven't even touched
> reiser4(resierfs 4.X).
>

ah, ok. From your post it sounded like you were using 4.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Mark Kirkwood]

Hemmann, Volker Armin wrote:
> On Saturday 28 October 2006 13:31, Mark Kirkwood wrote:
> 
>> I'd recommend changing to ext3 or xfs, as I've found both to be solid (I
>> prefer xfs but that just my personal opinion).
> 
> if you use XFS don't use 2.6.17 kernels.
> ...

A good link that briefly discusses power failures, write caching, 
kernels and write barriers is:

http://oss.sgi.com/projects/xfs/faq.html#nulls
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Richard Fish]

On 10/28/06, CapSel <capsel@gmail.com> wrote:
> 1. reiserfs breaks down on power failures even with option data=journal or
> sync

Huh, reiserfs doesn't do this.  It only logs metadata updates.

> 1. Is there the file system that preserves data & metadata like UFS2 ????
> "solid as rock!"

ext3 mounted with data=journal

> 2. What options do I need to enable (mount/mkfs) for XFS or JFS to enable
> FULL journaling?

Cannot be done.  They only log metadata.

> 4. Is there an option in kernel and how to enable it to sync on Oops and
> reboot?

Device Drivers
  ->Character Devices
    ->Watchdog Cards
      ->Watchdog Timer Support
        ->Software watchdog

You might also turn on the Magic SysRq stuff so you can do
"Alt-SysRq-s" to sync,
"Alt-SysRq-u" to remount everything read-only, and "Alt-SysRq-b" to
reboot the PC.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Richard Fish]

On 10/28/06, Uwe Thiem <uwix@iway.na> wrote:
> On 28 October 2006 13:14, Alexander Skwar wrote:
> > · Dale <dalek@exceedtech.net>:
> > > I'm even looking forward to Reiserfs4 myself.
> >
> > Well, actually I wouldn't hold my breath for Reiserfs anymore,
> > now that Hans Reiser is in jail. I really don't think that reiserfs
> > will have a good future.
>
> He isn't in jail but in detention. No court has convicted him yet.

Totally OT, but actually in the US, one normally goes to jail while
awaiting trial, and then prison if/when convicted.

-Richard

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Sunday 29 October 2006 01:45, Richard Fish wrote:
> On 10/28/06, CapSel <capsel@gmail.com> wrote:
> > 1. reiserfs breaks down on power failures even with option data=journal
> > or sync
>
> Huh, reiserfs doesn't do this.  It only logs metadata updates.

nope, it can log data too.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Richard Fish]

On 10/28/06, fire-eyes <sgtphou@fire-eyes.org> wrote:
> I'm not sure about the journaling, but the trick is that it caches
> writes very aggressively -- we're talking dozens of MB at a time, and it
> often holds onto those writes for not just minutes but even hours.

Just want to point out that XFS offers a few tuning parameters to
control how long it will cache metadata before writing it out.
Reducing these parameters "encourages" xfs to write associated file
data more frequently as well.

See /usr/src/linux/Documentation/filesystems/xfs.txt.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Richard Fish]

On 10/28/06, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de> wrote:
> On Sunday 29 October 2006 01:45, Richard Fish wrote:
> > On 10/28/06, CapSel <capsel@gmail.com> wrote:
> > > 1. reiserfs breaks down on power failures even with option data=journal
> > > or sync
> >
> > Huh, reiserfs doesn't do this.  It only logs metadata updates.
>
> nope, it can log data too.

Hmm, looks like "man mount" hasn't kept pace with the kernel.  Apologies.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Sunday 29 October 2006 02:33, Richard Fish wrote:
> On 10/28/06, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de> 
wrote:
> > On Sunday 29 October 2006 01:45, Richard Fish wrote:
> > > On 10/28/06, CapSel <capsel@gmail.com> wrote:
> > > > 1. reiserfs breaks down on power failures even with option
> > > > data=journal or sync
> > >
> > > Huh, reiserfs doesn't do this.  It only logs metadata updates.
> >
> > nope, it can log data too.
>
> Hmm, looks like "man mount" hasn't kept pace with the kernel.  Apologies.
>

yeah, that is a big problem. The data=journal option is AFAIR two years old.

The manpages are pretty... out of sync.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: BIG reiserfs problem

[Dale]

Richard Fish wrote:
> On 10/28/06, Uwe Thiem <uwix@iway.na> wrote:
>> On 28 October 2006 13:14, Alexander Skwar wrote:
>> > · Dale <dalek@exceedtech.net>:
>> > > I'm even looking forward to Reiserfs4 myself.
>> >
>> > Well, actually I wouldn't hold my breath for Reiserfs anymore,
>> > now that Hans Reiser is in jail. I really don't think that reiserfs
>> > will have a good future.
>>
>> He isn't in jail but in detention. No court has convicted him yet.
>
> Totally OT, but actually in the US, one normally goes to jail while
> awaiting trial, and then prison if/when convicted.
>
> -Richard
>

Well, depends on the crime and how much evidence.  If it is a serious
crime like murder and they have a lot of evidence, yup, that is true. 
It doesn't sound like they have a lot right now, evidence wise.  They
just have a serious crime, which is not good for him.

I just know that is a really pretty lady and I hope she is just off
hiding somewhere.  It doesn't sound good for him though.  They always
look at the spouse first, really hard.  It sounds like they were going
through a nasty divorce is why I'm just hoping she is off somewhere. 
That said, I doubt she would have left those kids behind if she did that.

My $0.02 worth.

Dale

:-)  :-)


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[William Kenworthy]

On Sat, 2006-10-28 at 20:25 +0100, Neil Bothwick wrote:
> On Sat, 28 Oct 2006 14:14:06 -0500, Joe Menola wrote:
> 
> > > > And DON'T use XFS if you can't afford an UPS.  
> > >
> > > Unless you're using a laptop.  
> > 
> > Solar UPS?
> 
> Battery!
> 
actually laptops are worse - on mine laptop-mode doesnt detect when a
battery runs out - so everything goes black!

On one occasion I lost most of an ext3 file system, so I went back to
reiserfs and no more problems.  I have lost small areas of data with
reiserfs (nothing recent though - very stable), but on ext3 Ive lost
whole systems (desktops, and the laptop mentioned above).

BillK
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Greg Bur]

On 10/29/06, William Kenworthy <billk@iinet.net.au> wrote:
> On Sat, 2006-10-28 at 20:25 +0100, Neil Bothwick wrote:
> > On Sat, 28 Oct 2006 14:14:06 -0500, Joe Menola wrote:
> >
> > > > > And DON'T use XFS if you can't afford an UPS.
> > > >
> > > > Unless you're using a laptop.
> > >
> > > Solar UPS?
> >
> > Battery!
> >
> actually laptops are worse - on mine laptop-mode doesnt detect when a
> battery runs out - so everything goes black!
>
> On one occasion I lost most of an ext3 file system, so I went back to
> reiserfs and no more problems.  I have lost small areas of data with
> reiserfs (nothing recent though - very stable), but on ext3 Ive lost
> whole systems (desktops, and the laptop mentioned above).

I'm a bit late joining the discussion but in my experience I would say
I had problems with reiserfs about 50% of the time spread over the
last six years or so.  Most of the problems were the result of power
failures but I had all sorts of strange things happen like
/etc/X11/X11.conf being replaced with /etc/profile or some such
nonsense.  Data loss happened occasionally as well.  I ended up
creating a cronjob that ran sync every five minutes and that pretty
much put a stop to the problems.. Now I avoid using reiserfs although
I do use reiser4 for a squid cache and there does seem to be a
noticable speed improvement.  Otherwise I'm slowly moving everything
over to jfs.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Chris Walters]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Joe Menola wrote:
> On Saturday 28 October 2006 2:02 pm, Neil Bothwick wrote:
>>> And DON'T use XFS if you can't afford an UPS.
>> Unless you're using a laptop.
> 
> Solar UPS?

Nuclear Reactor UPS
The good side, you can go a LONG time on UPS, and you can run your whole
house off from it...
The bad side, you have to pay to build the reactor and to dispose of the
waste (very expensive)...

LOL
Chris ;)
-----BEGIN PGP SIGNATURE-----

iD8DBQFFRG5zUx1jS/ORyCsRCDP2AJ9rZz5ulXHU4Gn6u6lEWBWYhCPwFQCgkUL8
laJWY7SKHpK9P1XnBbtHvVw=
=32Jd
-----END PGP SIGNATURE-----
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Novensiles divi Flamen]

[-- Attachment #1: Type: text/plain, Size: 484 bytes --]

On Sunday 29 October 2006 16:03, Chris Walters wrote:
>
> Nuclear Reactor UPS
> The good side, you can go a LONG time on UPS, and you can run your whole
> house off from it...
> The bad side, you have to pay to build the reactor and to dispose of the
> waste (very expensive)...

Too complex. Just get one with multiple power outputs. Then plug the UPS into 
itself. Much cheaper and safer. :p

- Noven
-- 
>-- Novensiles divi Flamen --<
>---- Miles Militis Fons ----<

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Benno Schulenberg]

Hemmann, Volker Armin wrote:
> On Sunday 29 October 2006 02:33, Richard Fish wrote:
> > Hmm, looks like "man mount" hasn't kept pace with the kernel. 
> > Apologies.
>
> yeah, that is a big problem. The data=journal option is AFAIR two
> years old.
>
> The manpages are pretty... out of sync.

This calls for submitting a patch.  But util-linux' development 
seems to have stagnated.  Adrian Bunk is listed as the current 
maintainer, but he appears to be busy with kernel 2.6.16 nowadays.  
And there doesn't seem to be any publically accessible cvs/svn/git 
repository from which to see how things are going.

Benno
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Sunday 29 October 2006 15:19, Benno Schulenberg wrote:
> Hemmann, Volker Armin wrote:
> > On Sunday 29 October 2006 02:33, Richard Fish wrote:
> > > Hmm, looks like "man mount" hasn't kept pace with the kernel.
> > > Apologies.
> >
> > yeah, that is a big problem. The data=journal option is AFAIR two
> > years old.
> >
> > The manpages are pretty... out of sync.
>
> This calls for submitting a patch.  But util-linux' development
> seems to have stagnated.  Adrian Bunk is listed as the current
> maintainer, but he appears to be busy with kernel 2.6.16 nowadays.
> And there doesn't seem to be any publically accessible cvs/svn/git
> repository from which to see how things are going.

yeah, luckily, there are 'current' mount options for reiserfs 3.6 on the 
namesys-hp.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

Dude - I use xfs w/o a UPS for desktops and laptops.  I use it on
servers with RAID and with UPS protection.  I also keep good backups
for the servers.   I have been using XFS since _just_ _after_ it came
to Linux.  I have used XFS on several hundred systems (which I have
been responsible for).   I have, to date, lost two filesystems.

 a. 2000 - I lost a filesystem when I was running a CVS kernel.
*hhahaha* yeah, it was ugly, during the 2.4 kernel of pain days.

 b. 2006 - hardware slowly corrupted an FS.  Some files wouldn't read,
and we had wierd problems, (but good backups).  After firmware updates
problems got stranger, and xfs_repair finished the job.   I blame the
hardware.

I have read the list, and seen the problems.  I don' t know what I do
that makes XFS succeed, but It really does work well.  The first
filesystem I ever tried with JFS failed.   I had weird errors, and
strange messages.   I tried the repair tools, but they crashed.  Then,
I posted to the LKML.  No one replied, or was interested.  I left JFS,
and returned to XFS.

I have run into a few strange bugs with XFS, but in every case I found
the mailing list and IRC very responsive and I was able to return the
servers to operation.   Twice those have been caused by either 2.4 or
XFS.   Once or twice it was several compound power outages.

What really kills XFS is _NOT_ power outages - it is out-of-order
commits.  When the drives re-order the commits, it really can f-up the
drive.  The data portion of the disk is updated and the journal isn't.
 Then, if you have a crash, you are in some pretty sh*t. That's why
write barriers are so important.

Use what you want, but don't misunderstand XFS - as many people here
clearly do.  It's a good FS, but it is sensitive to hardware problems.
 By problems I mean: dying disks (which will kill anyone), faulty
commit order for data vs. journal (which probably affects all of the
journaling FSs as well), silent corruption, faulty RAM, and last but
not least DMA problems.

If you have a drive that commits out of order, and you are prone to
power problems: USE EXT2 - it is , bar none, the SAFEST filesystem in
that case.   I do use it on a couple systems with those exact
problems. (And a couple of low-memory systems, journalling sucks up
resources).  I have not lost an ext2 fs yet in either of those cases.
Sure, the systems occasionally experience some meessed up files, but
never the whole FS, and replacing a library or binary is /not/ /that/
/tough/.

Good Luck,
  Joshua
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

On 10/28/06, Mark Kirkwood <markir@paradise.net.nz> wrote:
> Hemmann, Volker Armin wrote:
> > On Saturday 28 October 2006 13:31, Mark Kirkwood wrote:
> >
> >> I'd recommend changing to ext3 or xfs, as I've found both to be solid (I
> >> prefer xfs but that just my personal opinion).
> >
> > if you use XFS don't use 2.6.17 kernels.
> > ...

 2.6.17.8 and up work fine.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Hemmann, Volker Armin]

On Sunday 29 October 2006 16:56, Joshua Schmidlkofer wrote:
> On 10/28/06, Mark Kirkwood <markir@paradise.net.nz> wrote:
> > Hemmann, Volker Armin wrote:
> > > On Saturday 28 October 2006 13:31, Mark Kirkwood wrote:
> > >> I'd recommend changing to ext3 or xfs, as I've found both to be solid
> > >> (I prefer xfs but that just my personal opinion).
> > >
> > > if you use XFS don't use 2.6.17 kernels.
> > > ...
>
>  2.6.17.8 and up work fine.

and up to 2.6.17.5 it eats your data...
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

On 10/29/06, Hemmann, Volker Armin <volker.armin.hemmann@tu-clausthal.de> wrote:
> On Sunday 29 October 2006 16:56, Joshua Schmidlkofer wrote:
> > On 10/28/06, Mark Kirkwood <markir@paradise.net.nz> wrote:
> > > Hemmann, Volker Armin wrote:
> > > > On Saturday 28 October 2006 13:31, Mark Kirkwood wrote:
> > > >> I'd recommend changing to ext3 or xfs, as I've found both to be solid
> > > >> (I prefer xfs but that just my personal opinion).
> > > >
> > > > if you use XFS don't use 2.6.17 kernels.
> > > > ...
> >
> >  2.6.17.8 and up work fine.
>
> and up to 2.6.17.5 it eats your data...

That's not true.  Up to 2.6.17.5 it /may/ eat your data.  I ran
2.6.17, in the buggy state, for >2 weeks before I upgraded.  It did
_not_ eat my data.   Don't get me wrong - it was lucky, and yes - it's
a heinous bug.  If it just _ate_ your data, more people would have
lost data and or filesystems.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Norberto Bensa]

[-- Attachment #1: Type: text/plain, Size: 515 bytes --]

Novensiles divi Flamen wrote:
> On Sunday 29 October 2006 16:03, Chris Walters wrote:
> > Nuclear Reactor UPS
> > The good side, you can go a LONG time on UPS, and you can run your whole
> > house off from it...
> > The bad side, you have to pay to build the reactor and to dispose of the
> > waste (very expensive)...
>
> Too complex. Just get one with multiple power outputs. Then plug the UPS
> into itself. Much cheaper and safer. :p

Most probably it would just blow-up the UPS instantly :)

Regards,
Norberto

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] BIG reiserfs problem

[Chris Walters]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Joshua Schmidlkofer wrote:
> That's not true.  Up to 2.6.17.5 it /may/ eat your data.  I ran
> 2.6.17, in the buggy state, for >2 weeks before I upgraded.  It did
> _not_ eat my data.   Don't get me wrong - it was lucky, and yes - it's
> a heinous bug.  If it just _ate_ your data, more people would have
> lost data and or filesystems.

Well, I ran it then and didn't lose data or filesystems (very lucky),
but I switched to JFS because it was far faster on many things on my
system (such as deleting large numbers of files).  I guess everyone has
different experiences with these relatively new filesystems (XFS and JFS)...

Regards,
Chris
-----BEGIN PGP SIGNATURE-----

iD8DBQFFRNlIUx1jS/ORyCsRCL1CAKCPYF7T7phDAoLOaJwXFH7okvrYqwCdHbgX
hcotFCWcDVjLIu14lFMLWuY=
=yg4P
-----END PGP SIGNATURE-----
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

Yeah, I will try JFS again some day.  Not today, but  someday.   XFS
slow deletion is my personal pet peeve.   I have been using XFS for so
long, so successfully, that I am hesitant to change.  Plus, I have a
large number of existing installs.

Last year, I setup a JFS system on a dual-opteron.  I have no idea
what went wrong, but binaries wouldn't load, shit would crash, sig 11,
etc, etc, etc.   Finally, with now clue what was up, I reinstalled
using XFS.   There are a number of significant differences.   I am not
/blaming/ JFS, all I can say, it yet another JFS attempt went bad for
me.

So, XFS works, and I keep using it.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Alan McKinnon]

On Saturday 28 October 2006 19:51, Dale wrote:

> I'm thinking you could make the capacitors on the high voltage side
> MUCH larger so it would last longer.  It would still have to be a
> fast shutdown though.  Something like shutdown -h -t -5 minutes ago. 
> LOL The only thing about that is charging them when you first turn on
> the computer.

Your average psu uses around 5000 - 15000 uF caps on the mains side. 
That lasts about 3 mains cycles. I suppose you *could* rig up a 25F 
array but the caps will cost more than an average desktop machine (!)

Batteries = much cheaper.

The main reasons no vendor builds a UPS into the power supply:
1. Weight
2. Extra cost
3. Lead/acid battery = presence of sulphuric acid = bad news for 
motherboards

alan

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Uwe Thiem]

On 28 October 2006 23:39, Alan McKinnon wrote:
> On Saturday 28 October 2006 16:41, b.n. wrote:
> > Dale ha scritto:
> > > If you use XFS, make sure you have good power.  XFS does not like
> > > power failures at all.  I have had to reinstall on a second rig
> > > because of this very problem.  If you have a UPS, that may be OK.
> >
> > Thanks a lot for the advice. Power outages do happen and I don't have
> > an UPS. Why does it happen? Isn't XFS journaled?
>
> Yes it is journaled but it also allows data to be very aggressively
> cached. Make that VERY aggressively cached. With the result that data
> can be held in a huge cache somewhere and the kernel can be convinced
> it has been written to disk.

No journaled filesystem can 100% prevent data loss or even filesystem 
corruption in cases of power outages. Think of the the builtin caches of your 
drive. If that builtin cache contains a changed journal (not written to the 
actual drive yet) when a power failure occurs => bang!

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
http://www.SysEx.com.na
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Alan McKinnon]

On Monday 30 October 2006 12:04, Uwe Thiem wrote:
> On 28 October 2006 23:39, Alan McKinnon wrote:
> > On Saturday 28 October 2006 16:41, b.n. wrote:
> > > Dale ha scritto:
> > > > If you use XFS, make sure you have good power.  XFS does not
> > > > like power failures at all.  I have had to reinstall on a
> > > > second rig because of this very problem.  If you have a UPS,
> > > > that may be OK.
> > >
> > > Thanks a lot for the advice. Power outages do happen and I don't
> > > have an UPS. Why does it happen? Isn't XFS journaled?
> >
> > Yes it is journaled but it also allows data to be very aggressively
> > cached. Make that VERY aggressively cached. With the result that
> > data can be held in a huge cache somewhere and the kernel can be
> > convinced it has been written to disk.
>
> No journaled filesystem can 100% prevent data loss or even filesystem
> corruption in cases of power outages. Think of the the builtin caches
> of your drive. If that builtin cache contains a changed journal (not
> written to the actual drive yet) when a power failure occurs => bang!

All the more reason to take the intended usage of XFS seriously - in 
environments where power loss to the machine simply do not happen 
(redundant psus, UPS backup, etc)

alan
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Bryan Whitehead]

If you are so concerned with the awesomeness of XFS's caching... why not 
turn on data-journaling? Then data (not just meta-data) is committed to 
the journal.

You can also tune XFS to not wait so long to hold cached data.

Alan McKinnon wrote:
> On Saturday 28 October 2006 16:41, b.n. wrote:
>   
>> Dale ha scritto:
>>     
>>> If you use XFS, make sure you have good power.  XFS does not like
>>> power failures at all.  I have had to reinstall on a second rig
>>> because of this very problem.  If you have a UPS, that may be OK.
>>>       
>> Thanks a lot for the advice. Power outages do happen and I don't have
>> an UPS. Why does it happen? Isn't XFS journaled?
>>     
>
> Yes it is journaled but it also allows data to be very aggressively 
> cached. Make that VERY aggressively cached. With the result that data 
> can be held in a huge cache somewhere and the kernel can be convinced 
> it has been written to disk.
>
> Consider XFS's pedigree - SGI wrote it for their graphics machines. 
> These were big monsters backed up with high grade UPSs and such - the 
> logic was that if you spend a brazillion bucks on hardware, a mega UPS 
> is part of the deal, along with the wages to pay the army of admins you 
> also need.
>
> And, when doing video rendering, it turns out that it's easier to simply 
> re-render a frame when the filesystems does something odd with the data 
> rather than go to the effort of writing an FS that is 100% reliable. So 
> SGI sacrificed something that doesn't actually matter for their use 
> case to gain a significant performace increase (which does matter a 
> great deal)
>
> alan
>   

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Bryan Whitehead]

After sending this I realized that XFS doesn't support journal=data... I 
thought journal=data was a general VFS part of the linux kernel... my 
bad. :)

I guess you are just left with in kernel tuning (someone previously 
posted a link to).

Bryan Whitehead wrote:
> If you are so concerned with the awesomeness of XFS's caching... why 
> not turn on data-journaling? Then data (not just meta-data) is 
> committed to the journal.
>
> You can also tune XFS to not wait so long to hold cached data.
>
> Alan McKinnon wrote:
>> On Saturday 28 October 2006 16:41, b.n. wrote:
>>  
>>> Dale ha scritto:
>>>    
>>>> If you use XFS, make sure you have good power.  XFS does not like
>>>> power failures at all.  I have had to reinstall on a second rig
>>>> because of this very problem.  If you have a UPS, that may be OK.
>>>>       
>>> Thanks a lot for the advice. Power outages do happen and I don't have
>>> an UPS. Why does it happen? Isn't XFS journaled?
>>>     
>>
>> Yes it is journaled but it also allows data to be very aggressively 
>> cached. Make that VERY aggressively cached. With the result that data 
>> can be held in a huge cache somewhere and the kernel can be convinced 
>> it has been written to disk.
>>
>> Consider XFS's pedigree - SGI wrote it for their graphics machines. 
>> These were big monsters backed up with high grade UPSs and such - the 
>> logic was that if you spend a brazillion bucks on hardware, a mega 
>> UPS is part of the deal, along with the wages to pay the army of 
>> admins you also need.
>>
>> And, when doing video rendering, it turns out that it's easier to 
>> simply re-render a frame when the filesystems does something odd with 
>> the data rather than go to the effort of writing an FS that is 100% 
>> reliable. So SGI sacrificed something that doesn't actually matter 
>> for their use case to gain a significant performace increase (which 
>> does matter a great deal)
>>
>> alan
>>   
>

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

On 10/30/06, Bryan Whitehead <driver@megahappy.net> wrote:
> After sending this I realized that XFS doesn't support journal=data... I
> thought journal=data was a general VFS part of the linux kernel... my
> bad. :)
>
> I guess you are just left with in kernel tuning (someone previously
> posted a link to).
>

Well, even then you aren't guaranteeing "no data loss".
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Alan McKinnon]

On Monday 30 October 2006 22:49, Bryan Whitehead wrote:
> If you are so concerned with the awesomeness of XFS's caching... why
> not turn on data-journaling? Then data (not just meta-data) is
> committed to the journal.
>
> You can also tune XFS to not wait so long to hold cached data.

It's not something I'm really concerned about - I don't need the 
features XFS provides and ReiserFS 3.6 works just fine for me. I just 
understand why SGI built XFS the way they did and what problem they 
were solving by doing it that way.

I find it useful to keep in mind that XFS is a file-system (i.e. a 
system for files), and not necessarily a severly disk-bound filesystem

alan
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Uwe Thiem]

On 31 October 2006 09:17, Alan McKinnon wrote:

> I find it useful to keep in mind that XFS is a file-system (i.e. a
> system for files), and not necessarily a severly disk-bound filesystem

Would you mind to elaborate on this? I simply do not get your point.

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
http://www.SysEx.com.na
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Alan McKinnon]

On Tuesday 31 October 2006 11:04, Uwe Thiem wrote:
> On 31 October 2006 09:17, Alan McKinnon wrote:
> > I find it useful to keep in mind that XFS is a file-system (i.e. a
> > system for files), and not necessarily a severly disk-bound
> > filesystem
>
> Would you mind to elaborate on this? I simply do not get your point.

Historically SGI was very strong in graphics, and those applicatiosn 
tended to generate massive amounts of temporary files that had a short 
life and only the final version needs to be written to persistent 
storage, very well suited to aggressive caching and other similar 
speedups.

SGI's engineers could get away with this because they could guarantee 
that power loss to the machine wouldn't happen, so the potential data 
loss on a power outage didn't happen either. This sounds a bit weird to 
those of us raised on Intel where we pay close attention to getting 
everything on disk ASAP with as little performance loss as possible, 
but it's a perfectly reasonable system for an engineer to implement on 
the kind of hardware SGI were building.

That's why I say XFS is designed to not be tightly bound to the physical 
disk if the admin chooses to set it up that way, and the file system 
becomes more of a collection of directories and files that might never 
even be stored on a disk at all

alan
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Daniel Iliev]

Alan McKinnon wrote:
> On Tuesday 31 October 2006 11:04, Uwe Thiem wrote:
>   
>> On 31 October 2006 09:17, Alan McKinnon wrote:
>>     
>>> I find it useful to keep in mind that XFS is a file-system (i.e. a
>>> system for files), and not necessarily a severly disk-bound
>>> filesystem
>>>       
>> Would you mind to elaborate on this? I simply do not get your point.
>>     
>
> Historically SGI was very strong in graphics, and those applicatiosn 
> tended to generate massive amounts of temporary files that had a short 
> life and only the final version needs to be written to persistent 
> storage, very well suited to aggressive caching and other similar 
> speedups.
>
> SGI's engineers could get away with this because they could guarantee 
> that power loss to the machine wouldn't happen, so the potential data 
> loss on a power outage didn't happen either. This sounds a bit weird to 
> those of us raised on Intel where we pay close attention to getting 
> everything on disk ASAP with as little performance loss as possible, 
> but it's a perfectly reasonable system for an engineer to implement on 
> the kind of hardware SGI were building.
>
> That's why I say XFS is designed to not be tightly bound to the physical 
> disk if the admin chooses to set it up that way, and the file system 
> becomes more of a collection of directories and files that might never 
> even be stored on a disk at all
>
> alan
>   

"..we pay close attention to getting everything on disk ASAP with as little performance loss as possible.."

Then I would propose you to use "hdparm -W0 /dev/(what-ever)" to disable the write caching (no matter which FS you use). Nothing can give 100% guarantee against power failure.


-- 
Best regards,
Daniel


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

Daniels advice is actually the best that you can get.  It will give
you the smallest chance of corruption due out of order journal commits
that caching can cause.

js


On 10/31/06, Daniel Iliev <danny@ilievnet.com> wrote:
> Alan McKinnon wrote:
> > On Tuesday 31 October 2006 11:04, Uwe Thiem wrote:
> >
> >> On 31 October 2006 09:17, Alan McKinnon wrote:
> >>
> >>> I find it useful to keep in mind that XFS is a file-system (i.e. a
> >>> system for files), and not necessarily a severly disk-bound
> >>> filesystem
> >>>
> >> Would you mind to elaborate on this? I simply do not get your point.
> >>
> >
> > Historically SGI was very strong in graphics, and those applicatiosn
> > tended to generate massive amounts of temporary files that had a short
> > life and only the final version needs to be written to persistent
> > storage, very well suited to aggressive caching and other similar
> > speedups.
> >
> > SGI's engineers could get away with this because they could guarantee
> > that power loss to the machine wouldn't happen, so the potential data
> > loss on a power outage didn't happen either. This sounds a bit weird to
> > those of us raised on Intel where we pay close attention to getting
> > everything on disk ASAP with as little performance loss as possible,
> > but it's a perfectly reasonable system for an engineer to implement on
> > the kind of hardware SGI were building.
> >
> > That's why I say XFS is designed to not be tightly bound to the physical
> > disk if the admin chooses to set it up that way, and the file system
> > becomes more of a collection of directories and files that might never
> > even be stored on a disk at all
> >
> > alan
> >
>
> "..we pay close attention to getting everything on disk ASAP with as little performance loss as possible.."
>
> Then I would propose you to use "hdparm -W0 /dev/(what-ever)" to disable the write caching (no matter which FS you use). Nothing can give 100% guarantee against power failure.
>
>
> --
> Best regards,
> Daniel
>
>
> --
> gentoo-user@gentoo.org mailing list
>
>
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Ryan Tandy]

Etaoin Shrdlu wrote:
> On Tuesday 31 October 2006 18:50, Daniel Iliev wrote:
> 
>> Then I would propose you to use "hdparm -W0 /dev/(what-ever)" to
>> disable the write caching (no matter which FS you use). Nothing can
>> give 100% guarantee against power failure.
> 
> This disables only hard disk cache. If I understand correctly, the 
> caching done by the kernel (ie, by the filesystem driver) happens in 
> RAM, /before/ data is sent to disk. Did I miss something?

I think that's what Daniel intended.  If you were *really* paranoid 
about power outages, you could do that AND mount your FS with -o sync...
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Etaoin Shrdlu]

On Tuesday 31 October 2006 18:50, Daniel Iliev wrote:

> Then I would propose you to use "hdparm -W0 /dev/(what-ever)" to
> disable the write caching (no matter which FS you use). Nothing can
> give 100% guarantee against power failure.

This disables only hard disk cache. If I understand correctly, the 
caching done by the kernel (ie, by the filesystem driver) happens in 
RAM, /before/ data is sent to disk. Did I miss something?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Etaoin Shrdlu]

On Tuesday 31 October 2006 21:45, Ryan Tandy wrote:

> I think that's what Daniel intended.  If you were *really* paranoid
> about power outages, you could do that AND mount your FS with -o
> sync...

Ah ok, my misunderstanding then. Thank you!
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Mark Kirkwood]

Joshua Schmidlkofer wrote:
> Daniels advice is actually the best that you can get.  It will give
> you the smallest chance of corruption due out of order journal commits
> that caching can cause.

While this is true, it also may dramatically lower the mean time to 
failure for your disk, due to increased ware and tear - consumer ATA 
drives are designed to operate with the write cache on.

If you cannot afford to lose data due to poweroff corruption, then the 
only viable solution is a RAID card that includes battery backup (e.g. 
3Ware and Areca sell these).

regards

mark
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] BIG reiserfs problem

[Joshua Schmidlkofer]

> While this is true, it also may dramatically lower the mean time to
> failure for your disk, due to increased ware and tear - consumer ATA
> drives are designed to operate with the write cache on.
>
> If you cannot afford to lose data due to poweroff corruption, then the
> only viable solution is a RAID card that includes battery backup (e.g.
> 3Ware and Areca sell these).
>

Mark,

 Thank you for clarifying.  My comments were obscure and poorly
worded.  You are correct.  If one requires supreme data-safety with an
ATA drive, one must do things that will ultimately be
counter-productive for the longevity of your disks.   Furthermore, (I
can fully attest) that they will be a significant hit on performance.

  If you need high safety, speed, and retention, then the 3ware +
Battery backup is a low-cast solution.  There are tons of higher cost
ones.   most of me servers run with battery-backed RAID cards.

  I also like to make scripts to remount the FSs with -o sync in the
case of power outage.  That way, just in case the UPS actually fails
early (which happens at times), I don't lose data.   I have had that
preference no matter what FS i use.

sincerely,
  Joshua
-- 
gentoo-user@gentoo.org mailing list