Community:NDR/Journaling/LeaderProcedure

From NSDLWiki

Contents [hide] 1 Clean up the mess. 1.1 Be sure that the old leader is stopped. 1.2 Check each follower for open journal files. 1.3 Compare repository hashes on the followers. 1.4 Restart the RMI receiver on each follower 2 Modify the fedora.fcfg files 3 Start everything 3.1 RMI Recievers 3.2 Leader 3.3 Followers

Clean up the mess.

Be sure that the old leader is stopped.

• If the machine crashed, take it as read.
• Otherwise, stop tomcat:

/usr/local/tomcat-fedora/bin/shutdown.sh

Check each follower for open journal files.

This would be any journal file whose name begins with a hyphen:

ls /usr/local/ndr-content/journals/incomingFiles/_*

• If an open file is found, process it.

Rename the journal file to remove the leading hyphen:

cd /usr/local/ndr-content/journals/incomingFiles
mv _fedoraJournal20071023.051227.129Z fedoraJournal20071023.051227.129Z

Since the follower fedora is still running (if it isn't still running, start it), it will process the journal file normally.

Note: If the journal file ends abruptly, fedora may log an exception on the final transaction. However, all preceding transactions will have been processed, and the processed file will be moved to the archiveFiles directory.

Compare repository hashes on the followers.

TBD: How do we do this?

Restart the RMI receiver on each follower

This may be necessary when recovering from a crash. It should not be necessary if the leading fedora was shut down in an orderly manner.

/usr/local/fedora/RmiReceiver/stopReceiver.sh
/usr/local/fedora/RmiReceiver/startReceiver.sh

Modify the fedora.fcfg files

TBD

Start everything

RMI Recievers

TBD

Leader

TBD

Followers

TBD