Riak Search handoff bug in 1.2.0

Ryan Zezeski rzezeski at basho.com
Fri Oct 5 13:48:45 EDT 2012


Riak Users,

First, this bug is fixed in the upcoming 1.2.1 release [1].

There is a regression in Merge Index (the index engine for Riak Search) for
the 1.2.0 release that can cause search handoff to fail every time.  This
will result in a stuck cluster where new nodes never finish joining.  There
also seems to be a potential chain of events related to this bug that can
lead to corruption but after over a week of trying I wasn't able to
reproduce.  This bug requires that at least one of your Merge Index
segments be greater than the `segment_full_read_size` which defaults to
5MB.  The smoking gun to determine if you are hitting this bug is the
following log entry.

> [error] <0.1279.0>@mi_server:handle_info:549 Unexpected info {#Port
<0.123312424>,{data,[2,0,0,0,0,0,0,0,1|<<128>>]}}

If you are already on 1.2.0 and have hit this bug then you can upgrade to
1.2.1 and handoff will progress as it should.  However, if for some reason
your Merge Index data became corrupted then you'll need to address that
first before restarting the node.

-Z

[1]: https://github.com/basho/merge_index/pull/24
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20121005/62f344cb/attachment.html>


More information about the riak-users mailing list