We recently upgraded our Couchbase SDK from 2.4.6 to 2.5.0 and had to downgrade due to it introducing a memory leak. The attached image is a screenshot from New Relic Infrastructure of our six production web servers over the last 24 hours showing memory usage. We had to restart our node processes every four hours, hence the sharp declines, until we downgraded to 2.4.6 at around 1pm where we saw our memory usage stabilize. Hope this is the right place to share.
I am experiencing similar issue where the memory footprint would just keep on increasing with couchbase 2.5.0. On the other hand, couchbase 2.4.6 would work just fine.
We’ll have a look, sorry for the trouble!
Can you confirm that you use long-lived Couchbase bucket instance’s to communicate with the cluster (as opposed to creating them on the fly)?
That’s correct we use long lived couchbase bucket instances and only create new ones when they become disconnected.
I’ve identified the problem, filed an issue to track this and should have a fix available shortly. I plan to do our 2.5.1 release ahead of schedule (probably tomorrow) which will include the fix.
You can track the progress on the issue here:
Great news! Thanks for the quick turn around!