Usage of disk "/" on node is around 99%. core.* files

Hello all,

I am testing the following couchbase version: 4.5.0-2203 Enterprise Edition (build-2203).

Regardless of the reason, there is a particular process that seems to crash often. Every time this service crash, couchbase creates core files in the /opt/couchbase/var/lib/couchbase folder.

See here:
-rw------- 1 couchbase couchbase 1210572800 May 9 15:13 core.2295
-rw------- 1 couchbase couchbase 1584775168 May 9 15:14 core.2322
-rw------- 1 couchbase couchbase 1925439488 May 9 15:16 core.2345

1.2G core.2295
1.5G core.2322
1.8G core.2345

My / is currently set to 8GIG total, with avg fee space of 4.2GIG. As you can see, the space usage drop very quickly.

My issue is not the process crashing, but rather the core files generation space.

-Is there best practices I should follow regarding disk space with couchbase?
-Should I have a dedicated mount for something else than data and index ?
ex, should I have a dedicated mount for /opt/couchbase/var/lib/couchbase?
-Is there a max value I can set in couchbase to limit the space generated by core files?
-Is there a rotation process I should put in place? If so, which files should be rotated?
-What size I should dedicate for couchbase logs? as you can see above 4GIG is clearly not enough.

Thanks,

Steeve

Hi Steeve,

It sounds like you’ve hit a bug (crash) in the beta - first up it would be great if you could raise a bug on our issue tracker (issues.couchbase.com); ideally with a link to the core files if possible (as they should assist us in diagnosing the issue).

To your specific questions:

See the admin guide (http://developer.couchbase.com/documentation/server/4.5/install/installation-guide-intro.html) - in general you want to give each service (Data, Views, Indexes) which stores data it’s own partition for best performance.

Core file size is managed by the OS - look at the core ulimit in your OS documentation.

That is probably sufficient in normal use. Note that a core file (core dump) is a complete image of the process memory space, created upon an abnormal termination (segfault / abort), and as such isn’t expected in normal operation, so you shouldn’t expect to see lots of them.