Top Storage Consumers Report: Total Used Disk space is not accurate

[tjyang]

I have VM instance in miq version fine-3

• Ex1: Both used and provision are listed as 8.7TB.

• another place in miq to show ex1 VM's disk usage

• Ex1: in vcenter the total used disk space is 7.12TB, not 8.7TB(8.65TB).

• Ex2: miq report showing both are used and provisioned are at 4TB.
  

• Ex2: while from vcenter GUI, Used storage is only 1.76TB

[miq-bot]

This issue has been automatically marked as stale because it has not been updated for at least 6 months.

If you can still reproduce this issue on the current release or on master, please reply with all of the information you have about it in order to keep the issue open.

Thank you for all your contributions!

[tjyang]

I will check against Gapri-1 release to see if disk usage report is still not correct.

[agrare]

I think the issue here is that the :used_storage for a VM ends up using the disks' size_on_disk property and we can't get the actual used storage per disk from the VIM SDK. We can only get this per datastore at the VM level.

It would be possible to add vm-level used storage and override this core used_disk_storage method for vmware.

[tjyang]

Thanks for the hint Adam.

…

On Mon, Nov 26, 2018, 14:48 Adam Grare ***@***.*** wrote: I think the issue here is that the :used_storage for a VM ends up using the disks' size_on_disk property and we can't get the actual used storage per disk from the VIM SDK. We can only get this per datastore at the VM level. It would be possible to add vm-level used storage and override this core used_disk_storage method for vmware. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#98 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAXHDrXzd6imTz7pq2clG2vo1SbttLXLks5uzFOZgaJpZM4O214S> .

[paltaa]

Have this been fixed? or a workaround?

[agrare]

@paltaa I think smartstate might be able to get this data but we cannot get it from refresh, maybe @roliveri or @hsong-rh can help

[tjyang]

@paltaa , I reported this issue for vCenter provider.

[tjyang]

Hi All

I did another verification on this issue.

• On a hammer-2/vcenter 6.7 test environment.
• Before smartanalysis: zabbix VMware VM with thin 160G disk reported using 160G
• scan zabbix vm using smartanalysis
• space usage reported after smart analysis.
• Total diskspace report in dashboard widget is now reporting actual usage of 3G.
  

So this issue can be closed. VMWare VM actual space usage reported once it is scanned by smart analysis.
@paltaa , can you confirm and close this issue ?

[agrare]

Thanks @tjyang !

Yeah it should be more obvious that smartstate is required for these, we had issues with this recently with the snapshot size as well.

[tjyang]

@agrare , I think this issue happened even the VM got scanned by S.A. in earlier versions of MIQ.

[paltaa]

@tjyang Will test and confirm, thanks alot

[paltaa]

@tjyang Okay, so it was in fact a smart analysis problem, but since it creates snapshots of the VM's to get the analysis and the real disk usage, this creates a downtime of a few seconds that we cannot afford, so is there any other way we could get those metrics, or point me in the right direction to do it in the manageiq enviroment?

[agrare]

@paltaa we are unable to get file sizes from the VIM API which is used for refresh to smartstate is the only way to get this data.

[tjyang]

@paltaa ,

I don't believe the snapshot will be a few seconds long in a production VMWare setting that has hight end (flash) storage systems. I know netbackup/rubrik for VMWare are using snapshots for backup purpose. Unless your VM has very very high I/O otherwise taking snapshot should be fine. You can try S.A. on less busy VM first.

Also looks like you are using miq to manage production environment which is not recommended by miq team ;).

Also I voiced in another ticket that S.A. should not rely only on snapshot approach to gather system information. a simple ansible/ssh login shell script can collect most of the system info. Another point is that miq is is now managing phyiscal machines also.(no way to take snapshot).

Hope your question can be addressed by manageiq team members.

[agrare]

@tjyang @paltaa I don't see why we couldn't use datastore smartstate to get the file sizes since that uses the datastore browser, this would be a worthwhile RFE.

That would remove the need to use VM smartstate.

[paltaa]

@tjyang Thats the point, dont see the necesity to do a disk usage analysis by taking a snapshot if a small ansible playbook could get the metrics and return them to manageiq or any other system.

[tjyang]

@agrare

Do you want me to open up a RFE on S.A. without snaptshot action on providers ?

[paltaa]

@tjyang It would be nice, also is there a url to join that discussion/follow the thread or contribute to that solution

[agrare]

@paltaa well we wouldn't use an ansible playbook because we already have an efficient connection broker so we'd save sessions on your VC by not using ansible :)

@tjyang Yes I think that would be a good enhancement request, the description would be "Acquire disk usage with datastore smartstate not vm smartstate". It isn't going to be easy and will require a good bit of refactoring but there is a workaround.

Also @paltaa taking a snapshot for smartstate should not create "a few seconds of downtime". The only "downtime" is a momentary stun/unstun when coalescing the last of the writes into the base disk. If that takes longer than a tenth of a second there is something wrong. Snapshots are used on production VMs all the time, this is not unique to manageiq.

[tjyang]

@agrare, pls see HOW.3 at ManageIQ/manageiq-smartstate#84,
Let me know if you want me to have it as standalone ticket.

[agrare]

Hey @tjyang, I think this is a different issue because manageiq-smartstate is for scanning VM disk images where datastore smartstate uses the datastore browser.

I opened #374 to track this request.