You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@cassandra.apache.org by Fredrik Al <fr...@gmail.com> on 2016/03/02 11:10:11 UTC

Corrupt SSTables

Hi all.

Having two FSReadErrors:

FSReadError in
..\..\data\system\compaction_history-b4dbb7b4dc493fb5b3bfce6e434832ca\system-compaction_history-ka-329-CompressionInfo.db

FSReadError in
..\..\data\system\sstable_activity-5a1ff267ace03f128563cfae6103c65e\system-sstable_activity-ka-475-CompressionInfo.db

>From my understanding, the system tables are local (not replicated) which
means that removing those sstables and then run repair wont help.

If running scrub on those sstables doesn't help would it be safe to delete
those SSTables?

/Fred

Re: Corrupt SSTables

Posted by Alain RODRIGUEZ <ar...@gmail.com>.

Hi Fred,

Corrupted data due to software are quite rare nowadays. I would have a look
at the filesystem to first see if everything is ok. I recently had a case
where FS was unmounted and mounted back in a read only mode, Cassandra did
not like it.


   1. You should indeed give a try to:

   nodetool scrub system compaction_history

   2. If this is not working you can bring the node down and run the
   offline scrub


   https://docs.datastax.com/en/cassandra/2.1/cassandra/tools/toolsSSTableScrub_t.html

   3. "If running scrub on those sstables doesn't help would it be safe to
   delete those SSTables?"
   I am not sure about this one. I would probably go safely by dropping the
   node, cleaning it and bringing it back (using replace_address -->
   https://docs.datastax.com/en/cassandra/2.0/cassandra/operations/ops_replace_node_t.html).
   Or even remove the server if using a cloud, don't let instances bothering
   you if you're in the cloud, really...
   Yet I think this drop on sstable_activity + compaction_history is safe.
   I am just not sure about it.

>From my understanding, the system tables are local (not replicated) which
> means that removing those sstables and then run repair wont help.


Correct, you will not be able to run repair on the system keyspace due to
this (it makes no sense).

Hope this will help. If you find out that it is safe (or not) to remove
these sstables, I'll be happy to learn about that.

C*heers,
-----------------------
Alain Rodriguez - alain@thelastpickle.com
France

The Last Pickle - Apache Cassandra Consulting
http://www.thelastpickle.com



2016-03-02 11:10 GMT+01:00 Fredrik Al <fr...@gmail.com>:

> Hi all.
>
> Having two FSReadErrors:
>
> FSReadError in
> ..\..\data\system\compaction_history-b4dbb7b4dc493fb5b3bfce6e434832ca\system-compaction_history-ka-329-CompressionInfo.db
>
> FSReadError in
> ..\..\data\system\sstable_activity-5a1ff267ace03f128563cfae6103c65e\system-sstable_activity-ka-475-CompressionInfo.db
>
> From my understanding, the system tables are local (not replicated) which
> means that removing those sstables and then run repair wont help.
>
> If running scrub on those sstables doesn't help would it be safe to delete
> those SSTables?
>
> /Fred
>
>