Description
If a broker exits due to no space
<ioExceptionHandler>
<defaultIOExceptionHandler ignoreNoSpaceErrors="false"/>
</ioExceptionHandler>
i.e: some other process uses disk space. On a restart it will reset the limits and block pending space. Flow control will kick in. It may not be possible to consume messages in this stage.
However it will not allow clients to failover to another broker.
I think we should have an option to force the broker to error out, or fail to start if limits are set and they cannot be satisfied.
This will ensure that the broker will not accept connections in that state.
it could be
<broker adjustUsageLimits="false" ..>
When false, if the limits are exceeded, store usage or memory usage - the broker fails to start. When true, it adjusts them (like today) to what is available.
As ever, naming things is hard. Any better name?
Attachments
Issue Links
- relates to
-
AMQ-6134 ActiveMQ corrupted state when local disk fills up and Queues begin to buffer to KahaDB
- Closed