You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by John Sumsion <Su...@familysearch.org> on 2014/10/01 23:45:09 UTC
Question about incremental repair
If you only run incremental repairs, does that mean that bitrot will go undetected for already repaired sstables?
If so, is there any other process that will detect bitrot for all the repaired sstables other than full repair (or an unfortunate user)?
John...
NOTICE: This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message.
Re: Question about incremental repair
Posted by Robert Coli <rc...@eventbrite.com>.
On Wed, Oct 1, 2014 at 3:11 PM, Tyler Hobbs <ty...@datastax.com> wrote:
> Compressed SSTables store a checksum for every compressed block, which is
> checked each time the block is decompressed. I believe there's a ticket
> out there to add something similar for non-compressed SSTables.
>
> We also store the sha1 hash of SSTables in its own file on disk.
>
@OP : this came up a few weeks ago on the list, search for bitrot for the
previous thread.
Expanding on the discussion further, I plan to file a JIRA on
you-must-mark-all-sstables-for-that-range-unrepaired-if-you-fail-CRC-on-read.
I'll try to remember to reply on thread when I do.
Once there is a CRC on uncompressed read,
marking-all-sstables-unrepaired-on-failed-CRC would handle the bitrot case
for both uncompressed and compressed reads.
=Rob
Re: Question about incremental repair
Posted by Tyler Hobbs <ty...@datastax.com>.
Compressed SSTables store a checksum for every compressed block, which is
checked each time the block is decompressed. I believe there's a ticket
out there to add something similar for non-compressed SSTables.
We also store the sha1 hash of SSTables in its own file on disk.
On Wed, Oct 1, 2014 at 4:45 PM, John Sumsion <Su...@familysearch.org>
wrote:
> If you only run incremental repairs, does that mean that bitrot will go
> undetected for already repaired sstables?
>
> If so, is there any other process that will detect bitrot for all the
> repaired sstables other than full repair (or an unfortunate user)?
>
> John...
>
>
>
> NOTICE: This email message is for the sole use of the intended
> recipient(s) and may contain confidential and privileged information. Any
> unauthorized review, use, disclosure or distribution is prohibited. If you
> are not the intended recipient, please contact the sender by reply email
> and destroy all copies of the original message.
>
>
--
Tyler Hobbs
DataStax <http://datastax.com/>