[openstack-dev] Scheduler proposal

Clint Byrum clint at fewbar.com
Mon Oct 12 17:13:09 UTC 2015


Excerpts from Joshua Harlow's message of 2015-10-12 08:35:20 -0700:
> Thierry Carrez wrote:
> > Clint Byrum wrote:
> >> Excerpts from Joshua Harlow's message of 2015-10-10 17:43:40 -0700:
> >>> I'm curious is there any more detail about #1 below anywhere online?
> >>>
> >>> Does cassandra use some features of the JVM that the openJDK version
> >>> doesn't support? Something else?
> >> This about sums it up:
> >>
> >> https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/service/StartupChecks.java#L153-L155
> >>
> >>      // There is essentially no QA done on OpenJDK builds, and
> >>      // clusters running OpenJDK have seen many heap and load issues.
> >>      logger.warn("OpenJDK is not recommended. Please upgrade to the newest Oracle Java release");
> >
> > Or:
> > https://twitter.com/mipsytipsy/status/596697501991702528
> >
> > This is one of the reasons I'm generally negative about Java solutions
> > (Cassandra or Zookeeper): the free software JVM is still not on par with
> > the non-free one, so we indirectly force our users to use a non-free
> > dependency. I've been there before often enough to hear "did you
> > reproduce that bug under the {Sun,Oracle} JVM" quite a few times.
> 
> I'd be happy to 'fight' (and even fix) for any issues found with 
> zookeeper + openjdk if needed, that twitter posting hopefully ended up 
> in a bug being filed at https://issues.apache.org/jira/browse/ZOOKEEPER/ 
> and hopefully things getting fixed...
> 
> >
> > When the Java solution is the only solution for a problem space that
> > might still be a good trade-off (compared to reinventing the wheel for
> > example), but to share state or distribute locks, there are some pretty
> > good other options out there that don't suffer from the same fundamental
> > problem...
> >
> 
> IMHO it's the only 'mature' solution so far; but of course maturity is a 
> relative thing (look at the project age, version number of zookeeper vs 
> etcd, consul for a general idea around this); in general I'd really like 
> the TC and the foundation to help make the right decision here, because 
> this kind of choice affects the long-term future (and health) of 
> openstack as a whole (or I believe it does).
> 

Zookeeper sits in a very different space from Cassandra. I have had good
success with it on OpenJDK as well.

That said, we need to maybe go through some feature/risk matrices and
compare to etcd and Consul (this might be good to do as part of filling
out the DLM spec). The jvm issues goes away with both of those, but then
we get to deal Go issues.

Also, ZK has one other advantage over those: It is already in Debian and
Ubuntu, making access for developers much easier.



More information about the OpenStack-dev mailing list