cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Schuller <peter.schul...@infidyne.com>
Subject Re: Node forgets about most of its column families
Date Thu, 30 Aug 2012 04:16:09 GMT
> This is 1.1.X ?

Yeah. Post-1.1 release 1.1, a couple of months old w.r.t. upstream changes.

> Any thoughts on how recent the last schema change was ?

I believe a long time ago. Like weeks or more. Definitely not around
the time of the incident.

The first time I saw this happen after running scrub (triggered due to
upgrade from 0.8 -> 1.1 requiring scrub due to 0.8 bloom filter bug),
but later it happened on a regular restart on a post-upgraded system.

Haven't seen it on a cluster that hasn't been upgraded from 0.8, but
then I only saw it on one cluster so that is not a statistically
significant sample.

> Had the schema started in a pre 1.1X cluster? If so had their been a
> migration change after 1.1 upgrade?

Schema started on 0.8, no change after upgrade to 1.1. This triggered
on the very first node I ever ran scrub on in fact immediately after
the scrub (validation restart for paranoia after scrub), so there was
no opportunity to submit any schema change (due to mixed 0.8/1.1
cluster at the time).

-- 
/ Peter Schuller (@scode, http://worldmodscode.wordpress.com)

Mime
View raw message