Skip to contentSkip to conversation

reiserfs 3.6 problem with test9

DISCUSSION Samium Gromoff 3 participants 3 messages Nov 12, 2003

lore
Share & tools
lore
#0 lore raw
Quoted 6 lines
Quoted 3 lines

Last time I had a box with similar problems it was memory. I'd put your system through a memtest.

thanks for your answer.
I did as you said but no problem (memtest 3.0).

For how long did you run the test?

It is known that often a 24-hour memtest is not enough to find real world memory failures.

Also it is known that gcc (yeah youll wonder) is much better at finding them, so theres the question: do you compile stuff often, and if so, whether gcc bails out with "internal error" messages sometimes?

Also, if i were you i would do the following:

bash-2.05b$ dd if=/dev/urandom of=test bs=1048576 count=1024 bash-2.05b$ bzip2 test
bash-2.05b$ bunzip2 test.bz2

And if you have a memory failure bzip2 checksumming would show that up on the bunzip2 stage.

regards, Samium Gromoff

#1 lore raw
Quoted 17 lines · Samium Gromoff
Quoted 5 lines
Quoted 2 lines

Last time I had a box with similar problems it was memory. I'd put your system through a memtest.

thanks for your answer.
I did as you said but no problem (memtest 3.0).

For how long did you run the test?

It is known that often a 24-hour memtest is not enough to find real world memory failures.

Also it is known that gcc (yeah youll wonder) is much better at finding them, so theres the question: do you compile stuff often, and if so, whether gcc bails out with "internal error" messages sometimes?

Also, if i were you i would do the following:

Quoted 8 lines

bash-2.05b$ dd if=/dev/urandom of=test bs=1048576 count=1024 bash-2.05b$ bzip2 test
bash-2.05b$ bunzip2 test.bz2

And if you have a memory failure bzip2 checksumming would show that up on the bunzip2 stage.

regards, Samium Gromoff

thanks for your answer.
I ran memtest for a whole night yesterday (again) : nothing. I have never had any problem with gcc. execpt that sometimes after a big compiling I have some files damaged but _they_are_not_the_ones_used_ by_gcc_. they always are in another directory.
I ran the test with dd and bzip2 : no problem.

Yet I'm afraid it is too late. I switched to ext3 today as my all system was gradually becoming unusable after compiling large apps. I was fed up with rebuilding partitions.

Now, it runs fine. but it might not mean anything about my hardware ... I 'll try this.

Anyway I'm willing to explore the problem; in particular because some other people seem to have the same problem.
Moreover reiserfs is faster. I liked it and I'd like to come back to it.

Note: I kept the original reiserfs partition with the errors. So if someone needs any log to figure out where the problem comes from I have everything at hand. I'm also willing to perform other tests.

regards.
Philippe Rouquier.

#2 lore raw

Hello!

Philippe <[email protected]> wrote:
P> Note: I kept the original reiserfs partition with the errors. So if P> someone needs any log to figure out where the problem comes from I have P> everything at hand. I'm also willing to perform other tests.

Errors you've shown from your log suggest that somehow garbage appeared in some blocks that were supposed to be reiserfs metadata blocks. How this happened is unclear.
Also lots of people (including myself) use reiserfs on a daily basis with high loads (and on latest 2.6.0-test kernels, too) and have not seen anything like that so either there is something wrong with memory or with writing to your disk, I think.
Of course I might be wrong.
There are disk testing programs out there (sorry forgot the names) (not badblocks) that put a lot of load on disk and see how it does. May be you want to try one of those too.

Bye,
Oleg