Looking for screenshot from 6.4 or 6.5

[adamjg]

Hey, could someone do me a quick favor and go to a Windows VM object, then All Metrics, Guest File System stats, expand one of the letter drives and paste a screenshot of the metrics listed?  I think they removed a metric in 6.6.  Thanks!

[daphnissov]

vROps 6.3

vROps 6.5

vROps 6.6

[adamjg]

Ok this is weird and frustrating. It seems like for complex stuff vROps is great, but it lacks the ability to do simple day to day monitoring.

I'm looking to set up VM disk space alerts. I want to set a critical alert if a drive is 95% full and has less than 5GB free.  I had this set up and reporting correctly in a previous version (I want to say 6.4), but a bug in vROps was preventing this from alerting correctly. (example: an alert would trigger if say a C: drive was 95% full with 10GB free and a 5GB 😧 drive was 50% full).

There is no Guest File System Free (GB) for individual drives. I swear there used to be but based on your screenshots, they aren't there.  Not looking to use the endpoint agent at this time, as long as it still relies on java to run.

Any suggestions?

[daphnissov]

Little confused here. There is already an alert built for you out-of-the-box called "One or more virtual machine guest file systems are running out of disk space". It is built using symptoms "Guest file system space usage at critical level" and warning and immediate. The critical level fires when it reaches more than 95% utilization. So that part should already work with nothing you have to do. This phrase is not clear, however:

if a drive is 95% full and has less than 5GB free

Are you saying you want an AND condition to fire for both 95% and less than 5 GB free space available? Are you saying you want an alert that fires on a single condition being if remaining space is less than 5 GB?

The way the present alert is configured means it should fire if you have any drive above 95% (for critical), 90% (for immediate), and 85% (for warning). These can obviously be tweaked to your liking.

[adamjg]

Thanks for the reply. So those OOTB alerts only have the % free as you mentioned. What if you have a very large drive? If we're talking 95%, a 5TB drive will trigger critical at 250GB free.  What we've done is set 2 triggers, 95% AND less than 5GB free on the drive.  SCOM and SolarWinds do this OOTB very easily.

I guess an easier question is: what OOTB metric do I use to determine how much free space there is on an individual drive on a VM?

[daphnissov]

Ok, well it's just a matter of perspective. Most customers care about percentages rather than raw numbers because of their reflection of workload behavior. If you have a 5 TB drive, you're probably not filling it up with millions of text documents, and knowing when it only has 5 GB left (which is 0.001%, by the way), is probably going to completely fill before anyone has a chance to fix it. However, if that's what you want, you'll have to create a super metric that says "total disk capacity (GB) MINUS total disk used (GB)" and then use that resulting number to create both a symptom and then alert definition. Keep in mind about super metrics that they don't apply retroactively. So you'll have to create and then apply it to view data from that point forward.

[MichaelRyom]

HI adamjg

You are totally right it use to work! Havent seen a working solution since 6.2. Every version since would report wrongly based on multiple disk like you write later on in the thread.

My blog post here:

vRops Symptoms & Alerts - Disk usage - Michael Ryom

Shows the metric was there. Havnt check if you can enable it again in policy. But anyways it is a broken method.

Hope this answers your question.

daphnissov his reasoning is the same as your. He do not want false positive alerts and by using % together with absolute values you will be able to alert diffently on disk sizes wheahter they be 1g, 1t or 1p -byte.

There is a huge difference between 100mb or 100tb of free space. 100mb being to late maybe and 100tb most likely years in advance. At least for most customers.  

[daphnissov]

This is good info to know, MichaelRyom​. I did not know about that in 6.2. I do get the use case and understand it. So it seems like in post-6.2 versions you'll have to create a super metric, unfortunately. I would definitely let the vROps team know about this because I think it is quite useful so you can avoid having to use super metrics.

[adamjg]

Ok so I'm not losing my mind (at least not about this)!  Maybe a slight rant coming up...

We have different server groups that report on disk space differently.  Some VMs have disks that are always full but the apps have logs that rotate, so the disk should be full.  Some VMs we want to know the second we hit a threshold so the disk doesn't fill.  There's a lot of OOTB flexibility in other products, not so much in vROps.  The analytics are great in vROps, but some of those aren't intelligent enough to help.  Example: provision a VM with a 100GB 😧 drive that's empty.  Project gets put on hold, sits there for a month, then the app owner installs their stuff which includes 70GB worth of data.  vROps spits out "using space in an unexpected manner" and "projected to run out of disk space" alerts based on the trend.  It's a pain to get rid of them without just waiting another 30 days or whatever the trend is for them to settle down.

There are some OOTB alerts that aren't very helpful. For example: any alert on the total space of a VM with multiple disks is almost completely useless.  I could have a C: drive at 99% full, a 😧 drive at 0% full and vROps will think things are just fine. I have seen some additions in 6.6, but I haven't looked at this particular one yet.  Another one is Guest File System Usage. This seems like a really bad way to measure capacity, and a totally useless way to try and do actual monitoring. Guest File System Free is the metric we need.  As for why it was there in 6.2, I imagine they couldn't figure out the bugs in it so they fixed it as only VMware can, by removing it.

On the plus side, at least they didn't just change the documentation to say "don't do that" like they've done with other products!

[MichaelRyom]

Forgot to mention. Have talked with the vRops and from that I think native monitoring of disks is coming. But who knows - not me atleat    we can always hope for a major vRops announcement at VMworld  

[daphnissov · Accepted solution]

Native disk monitoring would be good, but I would be fine with them fixing the 6.6 Clarity release and putting back things that they tacitly removed from 6.5 and didn't tell anyone about.

[adamjg]

Seems like the super metric, besides being a band-aid approach, doesn't work with every individual drive. I'd have to create a super metric for each individual drive letter. I have a lot of VMs with 1 drive, but I have some with 20.

This thread got me what I was looking for, so I'm going to start another one to try and get the actual issue solved.  Thanks all!