VMware Cloud Community
dtracey
Expert
Expert

Alerting on non-OS filesystem usage on Linux

Hi Guys,


We have the EPOps agent deployed to a number of Red Hat 6 VMs, and have noticed that we are not seeing Alerts for non-OS file systems when they hit Warning, Immediate or Critical thresholds.

OS file systems (e.g /, /opt, /var, /home etc) all alert correctly, whereas custom file systems (e.g /applogs, /appdata) do not.

When drilling down through the OS object within vROps, I can see that the metrics are being gathered for each "FileServer Mount" (see screenshot), but these metrics are not being alerted on.

How do we configure vROps to alert on ALL linux filesystems (as it seems to do for Windows) current and future?  Or is it a case of changing the Metric / Property Symptom definition each time a VM comes along with a new file system?  This seems like unnecessary administration


Thanks,


Dan

filesystem.jpg

Reply
0 Kudos
6 Replies
AutomationGroup
VMware Employee
VMware Employee

NFS mounts are shown under RPC and not under FileServer Mount. Can you check if it is listed under RPC

Reply
0 Kudos
Urmilp
Contributor
Contributor

These are not NFS mounts....we have opened a call with VMware, hopefully they have a solution.

Reply
0 Kudos
AutomationGroup
VMware Employee
VMware Employee

I am from vROps R&D team. i am looking to check this inhouse.

It will be helpful if you can brief how you had created these custom filesystems and what you have mounted on them(Eg from your post -  /appdata and /applogs).

Can you also paste the output after running fstab command on the linux machine, for us to know what kind of a mount it is

Reply
0 Kudos
PaulHallettWTR
Contributor
Contributor

Output from fstab below.

/dev/mapper/rootvg-rootvg /                       ext4    defaults        1 1

/dev/mapper/rootvg-homelv /home             ext4    defaults,nodev        1 2

/dev/mapper/rootvg-perfloglv /perflog         ext4    defaults        1 2

/dev/mapper/rootvg-tmplv /tmp                 ext4    defaults,nodev,nosuid,noexec        1 2

/dev/mapper/rootvg-usrlv /usr                   ext4    defaults        1 2

/dev/mapper/rootvg-varlv /var                   ext4    defaults        1 2

/dev/mapper/rootvg-optlv /opt                   ext4    defaults        1 2

/dev/mapper/rootvg-swaplv swap             swap    defaults        0 0

/dev/mapper/datavg-wasappdatalv /wasappdata       ext4   defaults       0 0

/dev/mapper/datavg-optmqmlv /opt/mqm       ext4   defaults       0 0

/dev/mapper/datavg-varmqmlv /var/mqm       ext4   defaults       0 0

From a filesystem type perspective we have a combination of ext4 on RH6 and jfs on RH7.

(NB We have WAS and MQ installed)

Reply
0 Kudos
PaulHallettWTR
Contributor
Contributor

As a further note, the metric "FileServer Mount | Use Percent (%) " from the EP Ops adapter has no defined/built-in alerts in vROps.

However, the same metric from the Hyperic adapter DID have these alerts - and they were defined at the top level to capture ANY disk mounts current and future. (See below screenshot)

If we attempt to create a new alert definition for the equivalent EP Ops metric, we are only able to choose a specific disk mount and therefore would have to create one for each mount we have. Hopefully this makes sense!?!

Reply
0 Kudos
PaulHallettWTR
Contributor
Contributor

So to update this, I've managed to find a "workaround"

I created new symptom definitions, and chose the base object type of EP Ops Adapter -> Linux.

When choosing the metric to use, I selected FileServer Mount, and then a random mount point/utilization/use percent (%) metric as shown in Dan's previous screenshot.

Once I had selected this for the new symptom, I configured the name and static threshold level, but under the Advanced option I selected the "Evaluate on instanced metric" check box. The bubble tip had eluded to the fact that if I was to do this, it would use all instances of this metric. (I.E the metric for each of the available fileserver mounts)

By doing this, and then adding this symptom to a new alert we now received disk alerts for all mounts. Technically we could add the new symptoms to the existing alert, but we wanted to differentiate between the two. (At least initially)

It's not great, but it works. If you look at the symptom it still shows as being based on the specific instance of the metric I chose at random, but it does appear to work for all mounts across all VGs.

Reply
0 Kudos