You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by Jacob Rhoden <ja...@me.com> on 2014/11/19 03:30:16 UTC

Removing commit log files

Hi Guys,

Is it correct to assume that if you do a “nodetool drain” on a node and then shutdown a node, you can safely remove all commit logs on that node as long as all nodes are up? 

I have some VPS’s with low amounts of disk space that could do with it being recovered, I also assume this means startup time for that node will be drastically faster.

I have also experienced in the past, cases where if my tables had been altered, a restart would fail due to commit logs not being able to be replayed. I assume draining and removing the commit logs would be a good thing to do if you are worried about that bug occurring again (I don’t know if that bug was fixed or not).

Thanks,
Jacob

Re: Removing commit log files

Posted by Robert Coli <rc...@eventbrite.com>.
On Tue, Nov 18, 2014 at 6:30 PM, Jacob Rhoden <ja...@me.com> wrote:

> Is it correct to assume that if you do a “nodetool drain” on a node and
> then shutdown a node, you can safely remove all commit logs on that node as
> long as all nodes are up?
>

Assuming you are in a version where nodetool drain actually works, yes.
Most nodetool drain failures actually result in over-replay, not
under-flushing, so probably you are even ok in those versions.


> I have some VPS’s with low amounts of disk space that could do with it
> being recovered, I also assume this means startup time for that node will
> be drastically faster.
>

Yes, though you're trading shutdown time for startup time.


> I have also experienced in the past, cases where if my tables had been
> altered, a restart would fail due to commit logs not being able to be
> replayed. I assume draining and removing the commit logs would be a good
> thing to do if you are worried about that bug occurring again (I don’t know
> if that bug was fixed or not).
>

Yes.

=Rob
http://twitter.com/rcolidba