2013-09-03 02:00 AM
We are using NFS volumes as datastores for vSphere.
Volumes are managed via a dataset in UM.
They are thin provisionned with autogrow.
The volume nearly full threshold is set to 80 %.
The volume full threshold is set to 95%.
The vSphere admin configured alerts when datastore usage reaches 80%.
Example of what happen
Maximum volume space = 2 TB
Current allocated volume space = 1.7 TB
Current used space = 1.5 TB
Current % of used space from vSphere viewpoint = 88%
Current % of used space from UM viewpoint = 75%
Current % of used space from DOT = 88%
Many alerts are generated because from vSphere viewpoint the datastore usage is over 80%.
From UM viewpoint there is enough space because the current volume size does not reach the nearly full threshold.
From DOT viewpoint there is no need to autogrow the volume as it does not reach the autogrow threshold (~95%).
Do we have to avoid setting vSphere alerts on datastores as they are inconsistent ?
Do we better have to use UM alarms ?
Is there a way in UM to enforce volume growing up when used space reaches a given threshold of the current volume size ?
Many thanks for your help on this topic that could be part of best practices of vSphere on NetApp storage.
2014-01-28 03:43 AM
you make a very valid point.
I have exactly the same problem.
In OCUM 6.0 (Only available for Clustered ONTAP) they introduced a new feature:
Volumes with AutoGrow enabled will now be monitored by their MAX Autogrow Size instead of the current Vol Size.
So the current usage of the volume is matched against the max autogrowsize and not the current vol size.
This solves so many problems because now I get alot of events that a volume is nearly full or full.
BTW the trigger for ONTAP to autogrow the volumes are the following:
If the volume size is less than 20 GB, the autodelete threshold is 85%.
If the volume size is equal to or greater than 20 GB and less than 100 GB, the autodelete threshold is 90%.
If the volume size is equal to or greater than 100 GB and less than 500 GB, the autodelete threshold is 92%.
If the volume size is equal to or greater than 500 GB and less than 1 TB, the autodelete threshold is 95%.
If the volume size is equal to or greater than 1 TB, the autodelete threshold is 98%.
This makes it IMPOSSIBLE to monitor autogrow volumes with the current 5.2 OCUM.
I would really like to know from NetApp if they will implement the same fix as in OCUM 6.0 in OCUM 5.2 ?
2014-01-28 06:20 AM
Ah, one of my favorite topics, autogrow on NFS datastores with NFS...
First, let me address the Ontap autogrow settings. You can adjust when a volume autogrows via the registry
priv set diag;setflag wafl_reclaim_threshold_xl 90;priv set
That would need to be added to the filers rc file b/c it's a non-persistent setting
As for autogrow, 8.2 has autogrow / autoshrink which is a really cool feature which nees to be looked at if possible.
2014-03-21 03:50 AM
Thanks for your reply.
I am aware of the the registry settings but I don't want to change them.
I would like to change the way that OCUM monitors the volumes.
Like in OCUM 6.0, volumes are now monitored by their autogrow MAX size instead of the actual VOL size.
This way you will only get a volume (nearly) full message when the max autogrow size has been reached, this is an excellent solution!
About the new features in ONTAP 8.2 I know you can show set volume monitoring thresholds in ONTAP.
Do you know if they will be respected within OCUM?