[jira] [Comment Edited] (ARTEMIS-3291) It's possible to set pageSizeBytes and maxSizeBytes to zeroes, but it prevents broker from starting and unable to rollback

["Erikas Rudinskas (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-3291?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17342573#comment-17342573 ] 

Erikas Rudinskas edited comment on ARTEMIS-3291 at 5/11/21, 1:40 PM:
---------------------------------------------------------------------

For any poor soul looking how to fix it. In my case I had shared-store HA cluster (active-backup).

Steps to fix:
 # stop both Artemis services.
 # in master service's broker.xml, edit those 4 directories so they point to any other location in shared storage (in this example, from "data/journal" to "data/a/journal" where "a" is just any random subdirectory, stored in shared storage).
 # start master broker - it will not have any data and will be empty, but it is working.
 # delete "data/journal/server.lock", "data/journal/serverlock.1" and "data/journal/serverlock.2" files from shared storage.
 # create _hard_ links for the same 3 files - from "data/a/journal/server*" to "data/journal/" using below commands:
 ln data/a/journal/server.lock data/journal/server.lock
 ln data/a/journal/serverlock.1 data/journal/serverlock.1
 ln data/a/journal/serverlock.2 data/journal/serverlock.2
 # start slave Artemis service. At this point both web interfaces should be accessible.
 # kill (or stop) master Artemis service, so slave service becomes active.
 # using slave's "Web console --> broker (from tree view) --> Operations (tab) --> removeAddressSettings(String)" remove settings from broken address.
 # stop both Artemis services.
 # remove "data/a" directory from shared storage completely.
 # update master's broker.xml directories to point back to the old data.
 # start master server - it should be working fine. Also start backup server - it should also work fine.

Turns out backup server does not have the same validation as master server. This "issue" allowed us to fix this. :) 


was (Author: erikmnkl):
For any poor soul looking how to fix it. In my case I had shared-store HA cluster (active-backup).

Steps to fix:
 # stop both Artemis services.
 # in master service's broker.xml, edit those 4 directories so they point to any other location in shared storage (in this example, from "data/journal" to "data/a/journal" where "a" is just any random subdirectory, stored in shared storage).
 # start master broker - it will not have any data and will be empty, but it is working.
 # delete "data/journal/server.lock", "data/journal/serverlock.1" and "data/journal/serverlock.2" files from shared storage.
 # create _hard_ links for the same 3 files - from "data/a/journal/server*" to "data/journal/" using below commands:
ln data/a/journal/server.lock data/journal/server.lock
ln data/a/journal/serverlock.1 data/journal/serverlock.1
ln data/a/journal/serverlock.2 data/journal/serverlock.2
 # start slave Artemis service. At this point both web interfaces should be accessible.
 # kill (or stop) master Artemis service, so slave service becomes active.
 # using slave's "Web console --> broker (from tree view) --> Operations (tab) --> removeAddressSettings(String)" remove settings from broken address.
 # stop both Artemis services.
 # remove "data/a" directory from shared storage completely.
 # update master's broker.xml directories to point back to the old data.
 # start master server - it should be working fine. Also start backup server - it should also work fine.

> It's possible to set pageSizeBytes and maxSizeBytes to zeroes, but it prevents broker from starting and unable to rollback
> --------------------------------------------------------------------------------------------------------------------------
>
>                 Key: ARTEMIS-3291
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-3291
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>          Components: API, Broker, Configuration, JMX, Web Console
>    Affects Versions: 2.17.0
>            Reporter: Erikas Rudinskas
>            Priority: Major
>
> How to reproduce:
>  # Have perfectly fine working AMQ Artemis broker.
>  # Create *myAddress* address in any way you can.
>  # Open web form, in "Web console --> broker (from tree view) --> Operations (tab) --> addAddressSettings (first item in the list)".
>  # Add settings you want, but forget to update "pageSizeBytes" and "maxSizeBytes". These fields will be by default set to "0".
>  # Save settings.
>  # Restart broker and get the following error instead (broker is not starting):
> ERROR [org.apache.activemq.artemis.core.server] AMQ224000: Failure in initialisation: java.lang.IllegalStateException: java.lang.IllegalStateException: pageSize for address *myAddress* >= maxSize. Normally pageSize should be significantly smaller than maxSize, ms: 0 ps 0
>  
> It's impossible to fix this, because broker is not starting and values in broker.xml are being ignored...



--
This message was sent by Atlassian Jira
(v8.3.4#803005)