2010-05-11 02:39 AM
Hello expert gurus,
I need to monitor and NFS activity where my customer has millions of small files.
I need to define what counters and what values to set threshold on.
I was thinking of:
- NFS IOPS – no threshold
- NFS latency – thresholds: >10 ms: warning, >20 ms: alert
- CPU load % – thresholds: >80%: warning, >90%: alert
- Disk utilization % – thresholds: >80%: warning, >90%: alert
- Inode cache hits % – thresholds: <70%: warning
- Flash Cache hits % - no threshold
What do you think ?
Thanks for your help and comments.
Enterprise Systems Engineer
2010-05-11 02:53 AM
<<<< NFS latency – thresholds: >10 ms: warning, >20 ms: alert
AFAIK, Performance Advisor does not support multiple thresholds levels i.e. there will not be a warning and alert threshold value. Only two severities viz., error, normal are supported for a given threshold.
2010-05-15 01:40 PM
It's not specifically NFS related but I might also set a threshold for average disk busy on the aggregate in question.
You could also look at FlexShare to prioritize keeping metadata in cache over the actual file data.