[plug] Ubuntu woes

[Steve Baker]

I thought that too but wasn't sure.  I'm running 32bit version, 
linux-image-server kernel.  Just upgraded to 2.6.24-19-server, but this 
is no better (or worse) than earlier versions.

Write cache is enabled on the RAID adapter, I've had write performance 
problems with that setting before although it was never this bad.

Regards,
Steve


Paul Antoine wrote:
> Those write speeds are ridiculously low... there's no way it should be 
> that slow, even on RAID 5, especially given it's hardware RAID.
>
> Are you running the 64-bit version on your machine? The normal kernel 
> or the server version?
>
> Paul
>
> Steve Baker wrote:
>> Patrick Coleman wrote:
>>> You're not running LDAP on it, are you? I've seen similar issues with
>>> machines that are bound to LDAP (/etc/nsswitch.conf) but can't contact
>>> the LDAP server.
>>>
>>> The other thing to look for might be an NFS mount that can't contact 
>>> the server
>> Interesting thought, however - no.  No mention of ldap in 
>> nsswitch.conf, and this system doesn't have any NFS mounts on other 
>> servers (although it is an NFS server itself).
>>
>> I would think that the really slow boot (took nearly 20mins tonight 
>> when I restarted with upgraded kernel) is a key symptom.
>>
>> I checked the RAID during reboot, both arrays are optimal (neither 
>> are degraded).
>>
>> Timing for writing and then reading 100MB and 1GB files to each RAID 
>> are as follows (/tmp is on the RAID-1 and /home is on the RAID-5)
>>
>> (Apologies here for the long stream of data... summary: 
>> writing/reading 100MB and 1GB files, write speed 5.9-6.6 MB/s, read 
>> speed 60-116 MB/s)
>>
>> root at eclipse:/tmp# time (dd if=/dev/zero of=/tmp/testfile bs=1M 
>> count=100; sync)
>> 100+0 records in
>> 100+0 records out
>> 104857600 bytes (105 MB) copied, 16.0295 s, 6.5 MB/s
>>
>> real    0m17.397s
>> user    0m0.590s
>> sys    0m16.470s
>>
>> root at eclipse:/tmp# time (dd if=/dev/zero of=/home/testfile bs=1M 
>> count=100; sync)
>> 100+0 records in
>> 100+0 records out
>> 104857600 bytes (105 MB) copied, 17.6852 s, 5.9 MB/s
>>
>> real    0m19.172s
>> user    0m0.580s
>> sys    0m17.960s
>>
>> root at eclipse:/tmp# time (dd if=/dev/zero of=/tmp/testfile bs=1M 
>> count=1000; sync)
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 158.052 s, 6.6 MB/s
>>
>> real    2m39.904s
>> user    0m0.650s
>> sys    2m38.590s
>>
>> root at eclipse:/tmp# time (dd if=/dev/zero of=/home/testfile bs=1M 
>> count=1000; sync)
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 169.352 s, 6.2 MB/s
>>
>> real    2m51.243s
>> user    0m0.630s
>> sys    2m49.540s
>>
>> It's not uncommon for RAID-5 to be slightly slower on writes than 
>> RAID-1, at first glance these throughput rates don't look too far off 
>> what you'd expect for sustained transfer rates (for a single drive, 
>> anyway).  Is that right?
>>
>> Read times shown below are significantly better as you would expect.  
>> It's faster off the RAID-5 I guess because it is striped across 6 
>> drives instead of 2:
>>
>> root at eclipse:/tmp# echo 3 > /proc/sys/vm/drop_caches
>>
>> root at eclipse:/tmp# time dd if=/tmp/testfile of=/dev/null bs=1M count=100
>> 100+0 records in
>> 100+0 records out
>> 104857600 bytes (105 MB) copied, 1.73185 s, 60.5 MB/s
>>
>> real    0m2.454s
>> user    0m0.330s
>> sys    0m1.010s
>>
>> root at eclipse:/tmp# time dd if=/home/testfile of=/dev/null bs=1M 
>> count=100
>> 100+0 records in
>> 100+0 records out
>> 104857600 bytes (105 MB) copied, 1.18612 s, 88.4 MB/s
>>
>> real    0m1.563s
>> user    0m0.340s
>> sys    0m0.820s
>>
>> root at eclipse:/tmp# echo 3 > /proc/sys/vm/drop_caches
>>
>> root at eclipse:/tmp# time dd if=/tmp/testfile of=/dev/null bs=1M 
>> count=1000
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 12.7445 s, 82.3 MB/s
>>
>> real    0m13.107s
>> user    0m0.370s
>> sys    0m6.180s
>>
>> root at eclipse:/tmp# time dd if=/home/testfile of=/dev/null bs=1M 
>> count=1000
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 9.00243 s, 116 MB/s
>>
>> real    0m9.695s
>> user    0m0.380s
>> sys    0m6.320s
>>
>> Read performance seems to be fine, write speed seems a little slow 
>> but not enough to cause the sort of performance issues I seem to be 
>> experiencing.
>>
>> Does this give anyone else any other ideas?  I greatly appreciate all 
>> the tips received so far.
>>
>> Regards,
>> Steve
>> _______________________________________________
>> PLUG discussion list: plug at plug.org.au
>> http://www.plug.org.au/mailman/listinfo/plug
>> Committee e-mail: committee at plug.linux.org.au
> _______________________________________________
> PLUG discussion list: plug at plug.org.au
> http://www.plug.org.au/mailman/listinfo/plug
> Committee e-mail: committee at plug.linux.org.au
>