ignite-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Denis Magda <dma...@gridgain.com>
Subject Re: Fixed deadlock in GridDhtAtomicCache (Alex G. your review is needed)
Date Tue, 04 Aug 2015 19:48:48 GMT
Alex, thanks for the review!

Sure, this is just a local fix.
Recently I've detected and fixed several issues in TCP communication SPI
that happened because of invalidated cache context. In addition, Andrey
Gura mentioned that periodically he reproduces hangs in cache get
operations that most likely to happen because of invalidated cache context
as well.

Seems that it's time to fix the situation with invalidated cache context
globally. I'll create a task in JIRA in several days when return from a
short vacation putting extensive details. Then someone from the community
or me will have a chance to makes his/her hands dirty with this :)

As for this deadlock I'll merge that changes in any case because we need to
have them in the code to omit other RuntimeExceptions that may happen
because of any other reason. The threads that led to the deadlock were
threads from partitions supply pool or some internal workers pool.


On 4 авг. 2015 г., at 22:09, Alexey Goncharuk <alexey.goncharuk@gmail.com>

The change by itself looks right and can be merged, however I do not think
this is a complete fix. What kind of running threads were using invalidated
cache context? These threads may raise plenty of other exceptions if
invalid context is used. I think the proper solution should block a guard
(I am sure we already have a guard that we can reuse) and wait for all
threads to release this guard before cleaning up the context.

2015-08-04 8:28 GMT-07:00 Denis Magda <dmagda@gridgain.com>:

Hi Alex, Igniters,

I've fixed a deadlock in GridDhtAtomicCache that was a reason of frequent
hanging of "Cache Restart" test suite.

In short, the deadlock happened because a cache was already stopped but
some running threads, that perform cache related operations, keep using
invalidated GridCacheContext.
All the details are described here:
https://issues.apache.org/jira/browse/IGNITE-1189 <

Alex, as one of earlier implementers of this code, please review the


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message