Bucket compaction

[appajee]

Few questions on compaction behavior (with CB Server 5.1 & 5.5):

1.If we desire to kick off compaction manually, thru couchbase-cli – should the compaction be kicked off on “each node of the cluster” or just once from any node of the cluster?
2. What happens if a compaction op is kicked off, while a previous compaction is still running? Is it safe/non-intrusive OR must we ensure to check if a compaction is currently running before we kick off a new one?
3. If so, how to detect a. if compaction is currently running; b. the progress/stats of the currently running compaction job?

Thanks
AP

[househippo]

@appajee,

@appajee

1.If we desire to kick off compaction manually, thru couchbase-cli – should the compaction be kicked off on “each node of the cluster” or just once from any node of the cluster?

It does it for the whole cluster but the process picks a few vbuckets/shards/files per machine until it does them all. Here are the docs on compactions:

Auto-Compaction | Couchbase Docs

@appajee

2. What happens if a compaction op is kicked off, while a previous compaction is still running? Is it safe/non-intrusive OR must we ensure to check if a compaction is currently running before we kick off a new one?

In the below picture you can see when vbucket/shards/files get compact, even per node , in row 5 , column 3
docs fragmentation. Its just a ratio for row 5, column 1 vs row 5, column 2

@appajee

3. If so, how to detect a. if compaction is currently running; b. the progress/stats of the currently running compaction job?

Couchbase Is smart enough not to compact files that are currently being compacted.

Unless you have a write heavy use case I would recommend setting compaction to its defaults.