Couchbase stops ejecting keys. Returns nothing but Temp OOM errors until restarted.
I have a single node Couchbase 2.0 server, 16gb allocated to the cluster, with a bucket that expects ~650,000 new key/values a day. Each key/value around 100-200KB.
It seems though that after around 16GB of new writes, the server stops ejecting keys. The "memory used" goes beyond the low watermark (12.8GB) and beyond the high watermark as well (14.5GB) and never backs down.
I've tried tweaking the low and high water mark values. I've also set the min_data_age to 0.
Still with all this, it after about 16GB of new writes, restarting the server is the only thing that will make it work again.
Here is a screen-cap of the Bucket Analytics where the problem is visible.
My problem seems very similar to the one stated in the following link, but I don't think my issue involves metadata in RAM.
Any insight into what may be going wrong would be greatly appreciated. Also if I need to supply more information to help debug this, just let me know.