We’re facing a disk space issue on our Enterprise UM server host.
A single topic - LoyaltyTransactionTopic is consuming 101GB of disk space and its growing daily.
There is an auto maintenance set for the Topic and we’ve tried to purge All events in the UM Enterprise Manager but no luck.
Below is the stats:
[user@serverhost ]$ du -Sh | sort -rh | head -n 15
101G ./data/namedsubqueues/JMS/Kcms/LoyaltyTransactionTopic
23G ./data/namedsubqueues/JMS/Mdm/CustomerTopic
19G ./data/namedsubqueues/JMS/Common/AlertTopic
Attached is the All export config of the UM.
Could you anyone assist us in addressing this issue.
Please check the Realm property (in EM under Realms > Config):
Global Values > SharedDurableFilterBound.
This needs to be set to TRUE. If it is FALSE, then I would expect build-up of data in .mem files.
To activate this, you will need to do the following (which will result in the loss of any in-flight messages on these channels):
• Disable the JMS aliases in all ISes
• In EM, go to the relevant channel for the topic which has filtering, click the Named Objects tab, then click Get Names. Select and Delete the named objects that appear (one for each trigger)
• Re-enable the wM Messaging aliases. This will recreate the shared durables with the right settings.
You do indeed seem to have a lot of namedsubqueue files, which could point to some inconsistency in the internal queues. If the above actions don’t solve the disk space issue, then please open a Support Incident, as they can help you clean up the internal inconsistency.
Thanks, I could see the mentioned setting is already set to true.
Global Values > SharedDurableFilterBound=true.
I have opened a Support incident for this, awaiting resolution.
Additionally in EM under Realm Config > Event Storage
MaintenenceFileSizeThreshold is set to 2073741824
I’m assuming this setting will create a new .mem file once the Threshold value is reached. But in our case each .mem file in namedsubqueues are around 9GB.
Maintenance recovers space from files for messages that have already been subscribed. In your case, I suspect the files are orphaned shared durables containing messages that have not been subscribed (and marked for deletion) and hence will not cause files to shrink when maintenance is performed.
As you mentioned about orphaned messages above. Is there any way to delete these messages which has not been subscribed(marked for deletion) from the .mem file.
Issue we are having is almost same as in the original post where size of .mem file of a particular Topic(durable subscriber) is continuously increasing. Only difference to the original post above and our issue is that there is single .mem file under “namedsubqueues” folder.
UM version is 9.10. Auto Maintenance, SharedDurableFilterBound parameters are also enabled.
Arun,
the SharedDurableFilterBound is no longer present in recent UM versions, including 9.12 and it is configured to the correct behavior. So if you are seeing unexpected build-up of files on disk, then there is some other reason.
First check whether you have inactive subscribers, such as an IS trigger where the IS is no longer running, or where you have changed the client prefix of the IS. You can check the durable subscribers for a channel in Enterprise Manager and remove any that are obsolete.
If that is not the cause of build-up, then I suggest you open a support ticket.