mhh, i received the last answer via mail and did a reply via mail, but it seems 
that it did not get to this list.

so i repost via web:

--------------------------------------------------------------------------
> are you still having an issue with this? 

yes the issue still exists. 

even worse, meanwhile i had long-term stability issues with that system.
after one week of operation and several hundreds of gb of data transfer (read & 
write), the system stalled completely. 
this happend 3 or 4 times now. 
i could still ping it, but nothing more. no login from remote and no local 
login. 
before power-cycle on console there was some message from idmap "GSSAPI error - 
not specified GSS error....." 
so i run some process monitoring and after 2 days of operation, i suspect idmap 
is leaking memory, at least it looks so - vsz and rss of /usr/lib/idmapd grew 
by ~23mb, where vsz is growing almost linear.
ok, different issue - i have disabled the service, stopped the reboot and we 
will see if this is the root-cause of those hangs....
until today, this issue made me think that this is somewhat related to the read 
issue, but i think we have a completely different kind of problem here, as the 
system typically behaved like a box which is out of memory.

> it seems to me that the LSI 
> megaraid is having some issues with a higher queue depth .. by 
> limiting the device queue (vdev_max_pending) or file prefetch queue to 
> 1 you're making the problem go away

yes, i think it`s something like this. if it`s a problem very specific to this 
controller brand, i`m just ok with someone from sun telling "there`s nothing we 
can do, please contact the vendor", but i suspect a bug in the driver or even 
zfs layer.

> also the "page83 not standards compliant" message has to do with what 
> the vendor reports for the array controller (VPD = vendor product 
> data .. page83 tells us some useful information about multipathing) .. 
> do you have MPxIO enabled or no? This is going to push the problem 
> down to the sd device, but it seems like you're saying that reads here 
> are acceptable.
no, i don`t use MPxIO - at least i don`t know that i`m using it. it`s ordinary 
local raid5 storage (internal disks). 
to be 100% sure you may tell what i should check for.

thanks again

regards
roland
-- 
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org

Reply via email to