#Data ONTAP 8.1.2 disk util / cache 100% impact

1 messages · Page 1 of 1 (latest)

crisp shell
#

Hi Team.

My filer impact disk utilization 100%, Clients effect slow response.
I've check used for commands "::> sysstat"

What should i've check these problems?

aggr scrub /aggr1/plex0/rg1 is 46% complete (suspended)
aggr scrub /aggr0/plex0/rg0 is 25% complete
 
aggr scrub /aggr0/plex0/rg0 is 19% complete
aggr scrub /aggr1/plex0/rg0 is 9% complete (suspended)
aggr scrub /aggr1/plex0/rg1 is 16% complete (suspended)

Is this aggr scrub suspended will impact disk utilization?

Please let me know!!
Thanks and regards

quiet river
#

yeah, if the disks are at 100% there's not much you can do except remove workload from the system. Physical limits are physical limits ...
To answer your question, no, aggr scrub will not have any noticeable impact

crisp shell
#

Thanks for answer

quiet river
#

I don't know what ONTAP version that is but most recent versions should have statistics top file show and statistics top client show where you can find the worst offenders

crisp shell
#

I will try it! Thanks!

gray night
#

RAID scrubs on Sundays are fine.

crisp shell
#

@gray night

My filer still hit 100% Cache and Disk Util.

What aspects can we check to identify performance degradation that the client is experiencing and feeling uncomfortable with while using?

crisp shell
#

Wafl Is experiencing zombie throttling possibly due to requests for large number of file deletions. This can be mitigated by a combination of

a) reducing the load on the system

b) issuing the file deletion requests in smaller batches and

c) increasing the limits to allow more zombies to be queued on the system

#

This message can be issued with client performance?

gray night
#

Oh. Hmm.

#

Could be deletions, but I think wafltop works.

#

Obviously ignore the node run stuff.

#

Might give a clue.

crisp shell
#

How to stop wafltop works?

gray night
#

Should be in that KB...

crisp shell
# gray night Should be in that KB...

If i checked the performance with wafltop. Besides adding physical disks as Darkstar suggested, is there any other way to decisively solve this issue?

#

I plan to collect wafltop logs by running the following commands:

::> wafltop show -v cpu -i 10 -n 10
::> wafltop show -v io -i 10 -n 10
::> wafltop show -v latency -i 10 -n 10
::> wafltop show -v nvlog -i 10 -n 10
::*> wafltop show -v cputime -i 10 -n 10