2017-10-27 10:08 AM - edited 2017-10-27 10:09 AM
::system controller flash-cache*> system node external-cache show -node
Is Enabled: true
Is rewarm on: true
Is Mbuf inserts on: true
PCS size: 1024
Is hya caching enabled: true
The units in location A have a flashpool configured for specific aggregates, however the location B system currently does not have any Flashpools configured, so I strongly suspect that running in “Hybrid-Mode” is not the issue.
Looking at the hit-metrics, we’re not getting much value out of our FlashPools either, which I find hugely surprising, as our working set should comfortable fit into these, and our read times be massively accelerated by them – I can see that we’re getting really good hit rates (>75%) on our onboard cache – but that we’re still reaching all the way back to disk regularly. Is there some configuration option I have missed that will help us to shape read caching more optimally? Please see here:
Our write speeds are lovely
the ONTAP GUI interface repeatedly informs me that FlashCache is disabled – I can press the “enable” button, and see the optimization graph running, but if I browse away from the page and back it shows as “Disabled” again: is there a bug?
it says it’s “caching metadata only” (right underneath the unpleasantly high latency number) when enabled and disabled, which would seem a poor use for it given our entirely virtualised environment – but I don’t know how to change this to Default Mode (or even Low-Priority, although this is probably less suitable for our active data set).
2017-10-30 04:35 AM
We have the same problems at 2 customer sites:
FAS 8200 local cluster (Ontap 9.1) with 1 TB Flashcache per Head
GUI status of Flashcache is not correct.
Try using a local admin user (non Active-Directory User) for login in system manager
This is a bug in Ontap 9.1
Hit rates are below 5%
FAS 8200 fabric metro cluster (Ontap 9.1P8) with 2 TB Flashcache per Head
Hit rates of Flashcache are below 5%
Hit rates of memory are good