ESX storage utilization alarms

T

Thuleman

Supreme [H]ardness
Joined
Apr 13, 2004
Messages
5,833
I know that I read about this somewhere, I thought it was in the config guide, but I didn't see it in there after all.

Anyway, 75GB LUN on SAN, 74.75GB VMFS, 3.2GB free as reported in vClient. Alarm icon shows up for that datastore in the datastore view. No alarm shows up in the Alarms tab.

What's the rule as one goes from LUN space to creating a VMFS on that LUN, to utilizing that VMFS.
 
5% free is required (edit: in VMFS) - you may use the full LUN. I've had a snapshot chew up the free space - lets just say that the results were not pretty. I didn't have enough free space to delete even... Luckily it was in testing, as it would have resulted in serious downtime. I could only copy it to a larger datastore to fix (I think). I do remember what I ended up doing though, deleting the datastore and recreating it :(
 
Ugh, I must have missed that memo.
So I am not quite clear on how to fix this issue though. I can grow the LUN easily enough, but I can't grow the VMFS without to used extends which I don't want to use. Short downtime would be acceptable, but VM reinstall is not an option at this time.
 
vSphere, storage vmotion, cold migration.

10% free recommended for performance, 5% required for journal operations. :)
 
Another thing to look at is the hawt new migration options. It looks like you can pivot migrate thick vmdk's to thin vmdk's between datastores and back.
 
Alright, so in terms of storage though:

LUN size = VMDK size + 10%

OR

LUN size == VMDK size (minus some minor overhead) AND
VMDK size = partition size(s) + 10%

Where's the 10% coming in?
 
I have been doing a little more playing around with this and something isn't quite right yet.
100GB LUN created on SAN. SAN UI shows 80.1GB utilized, 19.5GB free.
Migrated VM over to new LUN. Windows shows drive C as 79.9GB.
That VM was 6GB RAM assigned. Storage alarm present in vCenter/vClient as little red triangle, no actual alarms present in the event viewer.

So ..., 80GB vmdk + vswp (let's just say 6GB) + 0.5 GB = 86.5GB * 1.1 = ~95GB ... on a 100GB LUN
Why then is it still upset about things?

How about this; Our math was wrong, we meant to be right, but we didn't write it up right.
We just assumed that LUN size = VMFS size? Which in the above is not the case. The VMFS size of the migrated box is still too small, hence the alarm? I didn't see an option to expand the VMFS to match the new raw LUN size. Am I just completely off track here?
 
When I create a datastore, I use the full size of the LUN. When the datastore becomes full, that's when you have problems.

Why is your SAN UI showing a 100GB LUN with only 80GB usable?
 
Yeah SpaceHonkey, right on, the key is to create the LUN, format it to VMFS using the "maximize capacity" option, and then it's necessary to not create a too large VMDK within that VMFS.

I migrated a formally 80GB LUN (fully used) to a 100GB LUN. The SAN UI shows 80GB used, not just 80GB usable. However, I am certain that when I created that 100GB LUN, I did format it to VMFS using the "maximize capacity" option. What I expected to happen was the VMDK end up in the now by roughly 20GB larger VMFS, and the alarm disappearing. That didn't quite happen. Looking forward to lopoetve's wisdom on this. ;)
 
yeah yeah... I'm 12 hours off of you all right now :p

Ok, so you migrated the lun - this won't expand the VMFS filesystem, as VMFS in 3.X doesn't support expansion (only extents) - thus, you have a 80gb filesystem on a 100gb lun, and 20gb of unused space (hence why the san is reporting that - most though won't recognize VMFS and will show either all used or all unused, no matter what you've done with it). :) Rebuild that VMFS, or build a new one, and move the vmdk to the new volume, and you'll be set!
 
It's like a never-ending story....

New 100GB LUN, formatted to VMFS for max capaciy.
New VM, 6GB RAM and thick 80GB HD assigned.

Datastore Warning icon shows up (yellow triangle)!
What the ..., why? Was I not paying attention?

(80GB + 6GB + 0.5GB) * 1.1 = 95.15
The datastore shows a capacity of 99.75 with 19.20 free in vSphere client.
 
4, still on the full blown eval since it won't accept my Advanced key to upgrade (have an SR in for that)
 
kk. I'll fire up vSphere when I get back from work this morning... probably early afternoon your time, and repro.
 
I did actually get to the bottom of this.
In vSphere there's a new alarm (probably more than one) added that by default triggers a storage warning (yellow) at 75% datastore utilization, and an alarm (red) at 85% utilization. I suppose this was added to make the admin more aware of renaming space on the datastore as it relates to thin provisioning.
 
OOH!

That's almost certainly it. I forgot about those... And you're right, they're for the thin provisioning option.

Sorry it took so long - I had to totally reinstall VC4, and doing so while using my VI3 in a box setup for training... yeah, talk about slow. :(
 
You must log in or register to reply here.
Back
Top