Garbage Collector

[priorcfl@gmail.com]

Does anyone have any knowledge of this process? I get that it is supposed to reclaim space, but what are you supposed to do when it stops doing its job? Seems to be the issue we are facing at the moment.

[bajorgensen]

It is a magical mystery to me. I have given up.
Just wait a week after UNMAP for the process to finish.

AFAIK there is a hidden task running, I do not think it is user interactive.
If anyone from HPE or partners who actually know, feel free to share!

- Anders

[MammaGutt]

Just throwing it out there, contact 3PAR support?

I've yet to find anywhere I can see the status of the process. If my memory serves me right you can do this in CLI to check what it has done (if running) for the last 24 hours

showeventlog -debug -oneline -msg GC -min 1440

[priorcfl@gmail.com]

Thanks for the replies. I had a elevated case open with HP and it eventually made it to the engineer level, our 3par actually dropped down to 3GB free space! We are still waiting for an official RCA.

[respection]

Have you got any answer from the support?
After unmapping the unused storage space I see very slow process of decreasing the reserved user space. And at the same the same time the result of the command compactcpg -dr shows nothing to compact which isn't understandable by me.

The only way I see to return unused space is to use command tunevv.
Am I right?

[sjm]

The garbage collection is a very slow process (in fact its capped)

How do I know this.....we had a 3par upgrade done by HP on one of our full SSD sans

They forgot to turn garbage collection back on......over the course of a few weeks we noticed it going down ....and logged calls.....it was not until I insisted that this started happening after the upgrade that an escalation engineer found this process off....Because its so slow...HP were forced to give us 4 SSD's for free or else our daily loads would have stopped the SAN in its tracks.

SSD's and garbage collection and using Dedupe on luns this use of space grows very quickly and needs to be reclaimed.

it took 3 to 4 weeks to get our space back...but we got to keep our new ssd's

[mrnie]

Scenario: After performing zeroing/unmapping of volumes, there is a large discrepancy between volume used space and volume reserved space so the released capacity is not available for other volumes and cannot be released/re-used in other CPG's.

Two ways to address this issue:

1) Wait for defrag to create contiguous space and release this capacity to the cpg (segments of 128MB - region stripe size). Defrag is a system internal process - there is no way to configure or influence it unless you have a root login. You can open a support case and they can configure defrag to run more frequently and in a more aggressive mode. They require to undo the settings afterwards despite no guarantee that it wont happen again and usually it still takes forever to release all the space as defrag takes one volume at a time and only limited segments of each volume for each run.

2) Volume conversion from thin -> thick and thick -> thin will defrag all the volume space and release it back to the cpg, this is usually faster than waiting for defrag + can be scripted. The downside is that it requires available space.

/Morten

[JeffWilson]

We have a 3par, completely ssd based with a similar problem, namely the "garbage collection" isn't working for 1 node and can't be restarted without rebooting all nodes and then applying a patch to stop the re-occurence.

In the meantime we are seeing a drop in space of between 500Gb - 1Tb per DAY. Obviously, we have only a limited time to apply the patch and reboots before the system fills up. At the moment we are estimating around 20 days to fill up.

Hopefully the restart will kick of the garbage collection process and start to reclaim space. Has anyone seen this process fail ? or put another way .....Will the 30 Tb the system has used over the past few weeks be reclaimed ?

We have to add SSD's to our systems 14 at a time so I doubt HP will be keen to offer free ones.

[Marc.mvh.vanhoof]

Could you please share your current 3PAR OS and the PATCH number ?

[JeffWilson]

Our 3Par is currently patched to:

Release version 3.2.2 (MU4)
Patches: P51,P56,P59,P84,P92,P94,P101

We can't go highter yet as the VM host baldes we have can randomly lose connection with out being upgraded to higher levels first

HP are proposing we apply patch 98. But this means losing 101 first I assume.