That's true, one has to avoid these exceptions. Setting limits is fine, and also allocating a separate partition for it. The remaining issue is "allow nothing else to write to it" (experience tells me Murphy will hit there). I now configured a broker plugin that frequently checks free disk space and stops broker connections when a given threshold is reached, starts them again when enough free space is available.
On 14.10.2015 14:38, Tim Bain wrote:
I'd say the tried and true approach to handling out-of-space exceptions is to prevent them from happening in the first place: set the storeLimit and tempLimit to appropriately large values, enable PFC to ensure you don't exceed them, and host them in a storage location that you can guarantee will always have enough space for both of them together. (For example, allocate a partition only for them that is 10% larger than the sum of the two limits, and allow nothing else to write to it.) Prevention is usually better than treatment. On Oct 14, 2015 4:36 AM, "Martin Lichtin" <lich...@yahoo.com.invalid> wrote: Wanted to ask if it is advisable to configure an IOExceptionHandler, such as <bean id="ioExceptionHandler" class="org.apache.activemq.util.DefaultIOExceptionHandler"> <property name="ignoreNoSpaceErrors"><value>false</value></property> <property name="stopStartConnectors"><value>true</value></property> </bean> I could not find much information about this feature, besides http://activemq.apache.org/configurable-ioexception-handling.html Is this a tried and trusted approach to handle e.g. an out-of-disk-space scenario? We've recently had a full disk and experienced lost messages, which is something we obviously try to avoid happening in the future. - Martin