Percona Link Mongodb Feedback

Wanted to give a bit of feedback to you guys on percona-link-mongodb. It’s really useful to migrate small/medium clusters (even cross region) and avoids having to deal with the whole backup and restore process.

I did however have to fork it in order to increase a timeout due to my cluster being quite large as it would otherwise fail on the 5 minutes hardcoded value.

It also acted a bit strange processing an oplog as it was only doing cycles of network burts every hour or so for 10-15min and then stopped without much disk or CPU activity which made it extra slow to catchup. This particular cluster does a lot of writes and has a TTL index doing auto deletes, which could be a reason why it acted that way.

Given this slow oplog processing which might be a bug or a quirk for this specific cluster, I didn’t end up trying it on my bigger (~1TB wiredtiger) clusters migrations and opted for a physical restore, but overall great tool!

Cheers!

@radoslaw.szulgo @daniil.bazhenov

4 Likes

Hey @Semantic

Really appreciate the feedback from you. Yes, currently replication performance is not what we are happy about and we have a task to improve it, we know what is the bottleneck. Hopefully in 0.8.0 we could have this improvement.

Regarding the timeout, I’ll create a task to make it configurable, that’l make it a bit easier to override rather than a fork :slight_smile:

Again, thanks for the feedback, it helps us a lot to make PLM better.

Cheers,

Inel

2 Likes

Glad that helped :smiley:

I could make a PR if the implementation is satisfying for you guys, no worries if not

1 Like

Please do, that would be awesome :slight_smile:

I believe your implementation is good, but I something comes up, we can discuss it in the PR.

Thanks!

1 Like