tree 08d023300753a8a66b10039b4146ae7847a33392
parent 16e8ba772bb5cf4c7546fb559bd9d455d4e42625
author Surinder Singh <sursingh@gmail.com> 1623892704 -0700
committer GitHub <noreply@github.com> 1623892704 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJgyqLgCRBK7hj4Ov3rIwAAb/cIAITxOpehzdqs+5qLzzXlJUIr
 GBKPhe6ZdcSa088vH6aAq9hJcka6AtZvSKrGOTpZyDT/VM1/dvLm3U/9LaGx+bV9
 aZoRJSpxvk478Rt7orK0moyqBZultXQ53WNFWPedAc6WacJxZuFcjtkM43SocDT4
 jESFcs6A8uZD9ZxxzGtgmBjBY50SRZDzUysWIhhrp6/hv8hKFdnMGNInOnSRK/YS
 WNKnhyE+slMa/i8XIJtG1RyokkUkKF7LfhrCeCjaC4lM1nK+68HzJKzCuHUcdSn0
 8mlC0rM0oUO5Azw6gPGITzQ1Q0XB70Rcv5j8AyIRzzVnDRJ7z9VqdyUPJjXbDec=
 =yElt
 -----END PGP SIGNATURE-----
 

[STATESTORE] Cleanup local storage for statestore storage container

### Motivation

While restoring a storage container, we fetch the checkpoint from the
checkpoint store. Currently this checkpoint will never get cleaned up. Every
time we restore the storage container on pod, a new checkpoint will get added.
Over period of time the disk usage keeps going up and eventually we have to
manually delete these stale checkpoints.

### Changes

With this change, we will cleanup the local storage for a storage container
whenever we close the KVStore. This will ensure that stale checkpoints are not
left behind. It is possible that POD may restart before the cleanup can be
done. To avoid these, we will also ensure that local storage for the storage
container is cleaned up before we restore the storage container.


Reviewers: Henry Saputra <hsaputra@apache.org>

This closes #2739 from sursingh/storage-cleanup and squashes the following commits:

6bd477691 [Surinder Singh] Clean local storage for storage containers
c74683695 [Surinder Singh] Add test case for local storage cleanup