[wildfly-dev] WFLY-818 Slave host does not register to DC
Brian Stansberry
brian.stansberry at redhat.com
Wed Jan 22 07:47:23 EST 2014
Agreed. The 3a) part needs some discussion; i.e. whether in this case
the slave just "asks" to restart or whether it just takes the update and
restarts servers. Currently if a slave loses contact with the master and
then reestablishes contact, it automatically restarts servers if the
domain config has changed. Basically we prioritize maintaining
consistency among members of a server group over giving humans control.
Plus, TBH, it's simpler.
If a server wasn't brought in sync automatically, I think it would have
to be put into a state where it could no longer receive or execute any
write ops. Reads are ok. Basically it's config is out of sync and it
needs to be restarted.
I'm not sure if there's any reason not to immediately apply the new
domain config to the HC itself, which is what we do now. Don't think
there is. If we didn't, the slave couldn't fully register with the
master, as all registered slaves need to be consistent with the master.
On 1/22/14, 5:58 AM, Emanuel Muckenhuber wrote:
>
> Hi,
>
> we don't have a solution yet. There are other open issues related to
> this, i.e. when using --cached-dc and --backup-dc at the same time. We
> do plan on allowing this case in future, which should behave similar to
> what you described. However it will probably only happen in the WidlFly
> 9 time frame.
>
> Emanuel
>
> On 01/22/2014 04:09 AM, Claudio Miranda wrote:
>> Hi, do you have any solution for this ?
>>
>> What do you think about this ?
>>
>> 1. HC starts with NO --cached-dc, it try to registers to DC, if it
>> doesn't register, HC fails to starts, as it is today.
>>
>> 2. HC starts with --cached-dc, it try to registers to DC,
>> a) if HC doesn't register to DC, HC uses the cached copy of
>> domain.xml. Try to registers to DC at interval times, if register is
>> successful see 3.a
>> b) if HC registers to DC, it works as is today copy domain.xml from DC
>> and replaces local backup.
>>
>> 3. HC starts with NO --cached-dc, HC register to DC, everything works
>> as is today. If HC disconnects from DC (unknown reason), HC keeps
>> trying to registers at interval times.
>> a) HC re-connects to DC, and detects domain.xml changed between DC and
>> HC. HC emits a warning and asks to restart the HC and their servers.
>>
>>
>> regards
>>
> _______________________________________________
> wildfly-dev mailing list
> wildfly-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/wildfly-dev
>
--
Brian Stansberry
Senior Principal Software Engineer
JBoss by Red Hat
More information about the wildfly-dev
mailing list