Feature request: Storage monitoring

[skruje]

I see that this was brought up several years back in post323.html but didn't find any update on it since.

I think it would be incredibly useful for Veeam ONE to be able to monitor storage systems, e.g. NetApp FAS, NetApp E-Series, Nimble. Some metrics/alarms that I would like to see are:

- Capacity, IOPS, throughput, and latency by volume, LUN, and NFS export with correlation to VMware datastore
- Storage jobs affecting performance, e.g. WAFL scans, SnapMirror, SIS
- Optimization of scheduled tasks on filer based on the above
- Unbalanced load on disks in aggregate (reallocation needed)
- Detection of bottlenecks: aggregates (add disks or reduce load), LUNs (split up datastores), network/storage interfaces (more links required)
- Storage health monitoring (back to back writes, high CPU load, controller failovers, temperatures)
- Tiering recommendations, e.g. move VM to SATA/SAS/SSD

Is this type of stuff on the roadmap or decidedly not in scope for Veeam ONE?

[Vitaliy S.]

Thanks for your feedback! Do you use any native/built-in tools for that right now? What storage vendor do you have right now?

[skruje]

We are using NetApp FAS and E-Series storage, so obviously those would be the priority for me For FAS, some of this functionality is in their Management Console software but I don't find it user-friendly at all. Most of the time I use manual commands sent to the console by ssh to get metrics, such as:

sysstat
stats start/show/stop
statit
wafl scan status
sis status
snapmirror status
reallocate status
aggr status -f
nfsstat

along with standard Linux text processing tools like grep/sort/cut/uniq, and the occasional perfstat, especially when working with support. For E-Series I haven't done much digging, because so far we are running only an EF550 (all flash) and never have any performance issues

It would be really great to wrap this stuff into Veeam ONE, which I am already using to monitor my VMware environment. It would be awesome to get advice when datastore latency is high during production hours that, for example, there's a dedupe scan running that could be deferred until later in the day. Or, as is sometimes the case, a physical server using a LUN on the same array, or a VM using an NFS export not exposed to VMware, is trashing performance for everything else. I think this kind of info could only be derived from polling the storage system directly.

You may also be able to detect VMs whose filesystems are not aligned to VMFS and the underlying storage and are causing IO amplification.

NetApp has an SDK / API that might help with this:
http://community.netapp.com/t5/Software ... td-p/97533

[Vitaliy S.]

These are good use cases, thanks for sharing. I will log this in our FR system for sure.