This post is more than 5 years old
70 Posts
1
7940
March 25th, 2015 17:00
Avamar 7.1.1 - Where has the ability to run Garbage Collection gone?
I have notice there is no option to run GC from the server tab in Avamar 7.1.1.
There doesn't seem to be a way to run it from cli either.
No Events found!
Avamar Exorcist
462 Posts
1
March 26th, 2015 01:00
Hi Bill
Good question.
Although the option to run garbage collection (GC) used to be present in the Administrator GUI it has not been useful for a long time.
Since Avamar v5 the first part of the GC process involves building a table of reference counts rather than actually deleting data. The 'GUI method' of running GC would set garbage collection to run with a timeout of 15 minutes.
If the time taken to build the table of reference counts exceeds this (which it will on most busy systems) then absolutely no data will get removed.
You can read more about how GC works (and used to work) in the following article.
124035 - Avamar space reclamation processes - Part 1: Garbage Collection
The removal of the feature from the GUI however is a good thing since running garbage collection manually is counter to best practices.
Garbage collection is intended to be an automated process and, providing the system doesn't completely fill up, it should be left to run naturally. If we decide to delete a lot of data from the system the maintenance windows can be extended to provide it more time to process the removed data.
One important point I'd like to make is that we should not need to micro-manage an Avamar system.
If there's a problem with a maintenance process (GC, checkpoint, HFScheck, crunching) then support should be engaged to investigate. We would never expect a user to need to run these manually unless there is a actual issue.
The only (rare) time when garbage might need to be run manually is where a system has completely filled up and has become read only
If an Administrator is following the guidelines in the following article this is very unlikely to happen.
KB 115872 - (Customer) Avamar User and OS Capacity Management
If a system does become read only, support should be engaged to help investigate the root cause of the capacity problem.
Only once the root cause of the problem has been identified and addressed might a L2 (senior support engineer) advise running manual garbage collection to tide the capacity over until the next automated run.
Hope that helps,
Nick
ae86levin
70 Posts
0
March 26th, 2015 01:00
Thank you for the detailed response
. That answers my question. The part about building the table in a 15min timeout certainly explains why I have seen GC fail to remove data that I knew it should be removing.
I frequently have to get one of our many Avamar grids/nodes out of admin mode.
Usually the case is that someone will dump 100's of GB of data onto a server without considering the repercussions.
So I will have to log a call any time this happens from now on - unless I can work out the commands to run it manually.
Avamar Exorcist
462 Posts
0
March 26th, 2015 02:00
Hi Bill,
You're very welcome.
I understand the concern about an "unexpected data increase" scenario...
One other problem to consider here is that you might also find that due to the sudden spike in new data the OS capacity may have increased beyond the point where the garbage collector is able to run (85% file system space). This would require some careful manual manipulation of checkpoints to reduce the checkpoint overhead before GC would be able to run naturally. This would be a task for support.
Unfortunately we've seen a lot of misuse of manual garbage over the years and since we want to avoid data loss at all costs we do err on the side of caution with regards to the procedures which are publicly documented.
Having said that, I know this is going to open a large can of worms but I'll contact engineering to discuss the possibility of an enhancement request that might allow a facility for a customer to self-run garbage collection in the event of the system having become read only.
Thanks,
Nick