bookkeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sijie Guo <guosi...@gmail.com>
Subject Re: Getting TOPIC_BUSY response
Date Sun, 06 Jan 2013 20:36:34 GMT
It seems that your subscription channel is broken and reconnect to hub
server again. but previous subscription state is not cleaned up, so you
encountered 'subscription already being served on a different channel'
issue.

but I could not figure out the root cause until you could provide more
details on following questions.

1) how many of bookie server used in your test? what is the ensemble size
that hub server used?
2) does your subscription disconnect from hub server? if it is possible,
could you explain more about your cluster setup?
3) after encountering this error messages, are you still receive any
published messages for the subscriber.

one guessing on this issue is related to
https://issues.apache.org/jira/browse/BOOKKEEPER-402 . when one bookie
server is down, a new published message found that there is not enough
bookies to persist that message. so it would release its topic ownership.
but in 4.1.0, releasing topic doesn't close subscription channel. so when
the subscriber reconnects, it would fail with the above issue. but it
doesn't make sense still: since in 4.1.0, we just support one channel per
subscription. so when it reconnects, it means the previous channel is
broken, which should cleanup the subscription state when a channel broken.
only very little chance to have this error message when race condition
happened as described in
https://issues.apache.org/jira/browse/BOOKKEEPER-513.

but it should be OK, since reconnect logic would try again to make the
subscription channel finally connected. so more infos are appreciate for us
to understand the case.

-Sijie


On Sun, Jan 6, 2013 at 8:25 AM, suman p <sumanp0309@gmail.com> wrote:

>
> Hi,
>
>  I continued playing with hedwig. I written a publisher and subscriber and
> started publishing some messages and receiving in subscriber. Everything
> worked fine for some time. After some time one of my book keeper server
> died, i restarted the book keeper server, then all of a sudden i am
> receiving following.I am not sure what went wrong. Can you shed some light
> on this. I am suing hedwig release-4.1.0
>
>
> Jan 5, 2013 7:23:28 AM org.apache.hedwig.client.netty.ResponseHandler
> channelDisconnected
> WARNING: Channel was disconnected to host: /72.37.13.106:4080
> Jan 5, 2013 7:23:28 AM
> org.apache.hedwig.client.handlers.SubscribeResponseHandler
> handleSubscribeResponse
> SEVERE: Unexpected error response from server for PubSubResponse:
> protocolVersion: VERSION_ONE
> statusCode: TOPIC_BUSY
> txnId: 3355
> statusMsg: "subscription for this topic, subscriberId is already being
> served on a different channel"
>
> Jan 5, 2013 7:23:28 AM
> org.apache.hedwig.client.handlers.SubscribeReconnectCallback operationFailed
> SEVERE: Subscribe reconnect failed with error: Server responded with a
> status code of: TOPIC_BUSY
>
>
> Regards,
> Suman
>
>

Mime
View raw message