bug with constant memory growth
I got a problem where memcached process would eagerly occupy all available memory (I caught it using 3.5 GB or RAM) and once you kill the process it starts again and tries to occupy all the memory again. Here I attach my log got from dump_logs.bat. The sequence was like this:
1. added new bucket (64MB).
2. deleted default bucket.
3. played some time setting server to Test mode (from Get Started section in server console) (had 0 of those test application buckets by the time when error appeared).
I got rid of this by:
1. removing my custom bucket (memcached process still occupied memory).
2. created bucket named "default".
Sorry guys, meant to respond to this earlier. There is a known issue in the 1.0.3 code where deleting the default bucket causes a memory leak. Re-adding it should stop the leak, but doesn't reclaim the memory so you'll have to restart the service.
Also, I would encourage you to try out the beta 4 release of our Membase server which includes support for standard memcached buckets and resolves this particular issue (as well as many others).
Thanks, let me know if there's anything else I can do to help.
Perry
I am having the same problem using NorthScale Memcached Server 1.0.3 on RedHat 5 Linux Platform. 2 identical servers setup as node to join the Cluster with a total of 8GB physical memory. A total of 5 buckets setup with a total of 6GB cache sizes. It only uses up to 1GB of cache sizes @ the moment. However, the console reports a total of 94% of memory being utilized from 50% last nite. ( an increases of 44% of memory usage overnite ?!) Those 2 boxes doesn't serve anything but memcache. This looks like a bug to me that there is some memory leak somewhere?