2010-10-28 11:43 AM
I have a client that knows that certain controllers will always have certain events always active; like volume full.
is there a way to have a controller ignore a single event or a class of events ( but still monitor for controller failure )?
2010-10-28 11:54 AM
In DFM --> Under "Setup" select "Alarms".
This will allow you to specify which events you would like to trigger alarms for and also which groups should be monitored for these events. You could put the controller you know will have frequent alerts in it's own group and exclude it from "volume full" event alarms.
Hope this helps.
2010-10-28 12:07 PM
There's a couple of possible techniques.
You could adjust the thresholds for the volumes in question to keep the event from firing. Most thresholds can be set at both the global and object level for this reason.
You could change the event severity for the event type in question. For example, if you didn't want to see any volume nearly full events, you could use this:
$ dfm eventtype modify -v Normal volume-almost-full
The event will still be generated but it won't change the volume or filer resource status.
Finally, you could just stop the monitor in question. If you set the monitor interval to 0, we will not execute that monitor. So if the events are all coming from, say, the vFiler monitor, and you never, ever will care about vfilers, use:
$ dfm option set vfilerMonInterval=0
2010-10-28 12:20 PM
Adding to Pete's point:
Adjusting the thresholds of the volumes in question is best way - Reason: It's specific to induvidual objects.
You can adjust the severity of an event/s or adjust the monitoring interval, however these are global settings - which means it would affect the entire set of objects monitored by DFM.
Hope that helps
Thanks and regards
2010-10-28 12:25 PM
Would you mind explaining how you would define a threshold for specific volume (If that is in fact what you're proposing, perhaps I've misunderstood)?
I'd thought it was only possible to define them globally through "Setup" --> "Options" then selecting "Default Thresholds" and then specifying the "Volume Full Threshold (%)" & "Volume Nearly Full Threshold (%)" values.
2010-10-28 12:38 PM
Here are the steps:
You need to get into the volume detail's page (Just click on a volume in Operations Manager and you will be taken to the volume details page). Once in, look out for something call "Edit Quota Settings" on the left bottom of your window (Check the screenshot). Click on it and you should be able to see the threshold settings for that particular volume.
type the command: "dfm volume set" and you should be able see the following options:
C:\Users\Administrator>dfm volume set
Valid options are
volFullThreshold Volume Full Threshold (%)
volNearlyFullThreshold Volume Nearly Full Threshold (%)
volFullThresholdInterval Volume Full Threshold Interval
userEnableAlerts Enable User Quota Alerts, not available for vserver volumes
userFullThreshold User Quota Full Threshold (%), not available for vserver volumes
userNearlyFullThreshold User Quota Nearly Full Threshold (%), not available for vserver volumes
volSnapshotFullThreshold Volume Snap Reserve Full Threshold(%)
volNearlyNoFirstSnapThreshold Volume Nearly No First Snapshot Threshold(%), not available for vserver volumes
volNoFirstSnapThreshold Volume No First Snapshot Threshold(%), not available for vserver volumes
volReserveNearlyDepletedThreshold Volume Space Reserve Nearly Depleted Threshold(%)
volReserveDepletedThreshold Volume Space Reserve Depleted Threshold(%)
volGrowthEventMinChangePct Volume Growth Event Minimum Change (%)
volNearlyOvercommittedThreshold Volume Quota Nearly Overcommitted Threshold (%)
volOvercommittedThreshold Volume Quota Overcommitted Threshold (%)
volSnapshotCountThreshold Volume Snapshot Count Threshold
volSnapshotTooOldThreshold Volume Too Old Snapshot Threshold
volNearlyOverDeduplicatedThreshold Volume Nearly Over Deduplicated Threshold (%)
volOverDeduplicatedThreshold Volume Over Deduplicated Threshold (%)
C:\Users\Administrator>dfm volume set help
set -- change volNearlyFullThreshold and volFullThreshold
dfm volume set <object> <option-name>=<option-value> ...
Changes one or more options of a volume.
All option names will be listed by executing
this command without any parameters.
I hope that helps.
Thanks and regards
2010-10-28 07:12 PM
Another way not to get the events for the specified objects are delete the object from dfm using the delete commands.
Like, if you know a controller is down for tech refresh for, say a week, you can stop getting alerts from this controller by just doing the below.
dfm filer delete <filername-or-id-ip>.By that way its only marked deleted in the dfm database and not completely removed.
After one week if you would like to monitor it again, dfm filer add <filername-or-ip>.
This can be done for any objects like, volume qtree, aggr,lun etc.
2010-10-28 07:20 PM
That's a good point (@ adai) - It actually gives a pretty good idea of how object-deletion (it's not deletion but just removal of an object) works in Operations Manager you can undelete (add the object back) the object if required - that's kinda cool.
However do keep in mind that if you remove the filer - The entire filer will not be monitored untill you add it back - which means that you won't recieve any alerts on that filer - as opposed to just avoiding volume-full messages on unimportant volumes.
Thanks and regards
2010-10-28 07:43 PM
Yes, shiva's correct, When you delete a object say a filer, all aggr, volume, qtree, and lun(basically its child objects) are not monitored any more untill they are un-deleted(re-added).
If your intension is only to avoid a specific event they go with changing the threshold setting for individual objects.
If your intension is to not to monitor the entire volume for any event then go with delete.Also keep in mind when we mark it as deleted, all the child objects are also automatically marked deleted and not monitored anymore.
Like if a volue is marked deleted, its qtrees and luns are also not monitored anymore.
similarly if its a aggr, all its volume, qtree and lun are not monitored.