activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "怀守昆" <huai...@sina.com>
Subject 转发:Re: Re: Artemis backup failover even quorum vote failed because of empty networkhealthcheck
Date Wed, 10 May 2017 07:03:27 GMT
Clebert,

I see Justin's PR#1260 fixed empty network health check problem.

Can you consider another problem:

why live stop replication before quorum vote result come out?in this case, if vote result
is to shutdown, then we my possibly lose some message.

I tested when live stop replication, before vote, live is still serving messages.

May be we can expose storageManagerLock in JournalStorageManager, so we can lock the journal
before vote completed.


----- 原始邮件 -----
发件人:Clebert Suconic <clebert.suconic@gmail.com>
收件人:users@activemq.apache.org, huaishk@sina.com
主题:Re: Re: Artemis backup failover even quorum vote failed because of empty networkhealthcheck
日期:2017年05月05日 21点49分


I'm not sure how you're doing this...
you should have 3 pairs of active/backup for this to work...
or if you don't care about having slaves for 2 of the servers?
it seems to me that you have shared storage and 2 servers waiting to
become live? they won't be part of the topology until they get the
lock.
I will need some clarification and version on what you're doing.
On Fri, May 5, 2017 at 1:16 AM, 怀守昆 <huaishk@sina.com> wrote:
> Sorry for late response.
>
> Just updated test plan, with 3 master and 1 slave, using static connectors.
>
> After 4 node started, the slave become backup of 1 master.
> Then i use iptables to block outgoing packet on port 61616 at slave side, so it is kind
of isolation.
> 60 seconds later, backup become live with the same node id.
>
> I tried master and the "slave", both can accept message.
>
> One more question, why live stop replication before quorum vote result come out?
> in this case, if vote result is to shutdown, then we my possibly lose some message.
>
>
> ----- 原始邮件 -----
> 发件人:Clebert Suconic <clebert.suconic@gmail.com>
> 收件人:users@activemq.apache.org, huaishk@sina.com
> 主题:Re: Artemis backup failover even quorum vote failed because of empty networkhealthcheck
> 日期:2017年05月05日 10点57分
>
>
> can you try master?
> On Thu, May 4, 2017 at 10:53 PM, 怀守昆 <huaishk@sina.com> wrote:
>>  I'm test Artemis network isolation using quorum vote.
>>
>> The backup failover after quorum vote return false.
>>
>> I check the code find that in SharedNothingBackupQuorum#decideOnAction, after test
isLiveDown(), networkHealthCheck is also checked.
>> If no network health check url/address configured, it always returns true.
>> So the backup decide it should failover.
>>
>> Will this be an issue?
>>
>> I use artemis 2.0.0, with 1 master and 1 slave.
>> According user guide, the backup should never failover.
> --
> Clebert Suconic
-- 
Clebert Suconic
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message