Running a full lock re-index prevents incoming mail processing on a node

XMLWordPrintable

    • Type: Bug
    • Resolution: Unresolved
    • Priority: Medium
    • None
    • Affects Version/s: 9.4.1, 9.5.1
    • Component/s: Indexing, Scheduled Tasks
    • None
    • 9.04
    • 1
    • Severity 2 - Major

      Issue Summary

      The incoming mail processing service for Jira by default runs every 1 minute. Whenever a full lock re-index is run on a node, this prevents the incoming mail processing service from running. The effect of this behaviour on single node Jira instances is that any incoming mail processing is halted until the full lock re-index is finished. For multi-node instances, if the incoming mail processing service is scheduled for a node that has the full lock re-index running, this means that processing incoming mails are impacted until the full lock re-index is done.

      This is reproducible on Data Center: (yes)

      Steps to Reproduce

      1. Configure an incoming mail handler.
      2. Run a full re-index.
      3. Send several mails to your mailbox configured for the mail handler.

      Expected Results

      For multi node set-ups, the incoming mail processing service should not run on the node running the full re-index.
      For single node set-up: Warning or documentation that any incoming mail processing will be paused until the full re-index completes.

      Actual Results

      MailFetcherService normally runs per minute by default. However, when the full reindex is triggered, the MailFetcherService is not called until the full reindex completes.

      atlassian-jira-incoming-mail.log.2:2023-01-20 10:24:00,010+1100 DEBUG [] Caesium-1-4 anonymous    breaklotsofstuff MailFetcherService run() method has been called
atlassian-jira-incoming-mail.log.2:2023-01-20 10:25:00,011+1100 DEBUG [] Caesium-1-2 anonymous    breaklotsofstuff MailFetcherService run() method has been called
atlassian-jira-incoming-mail.log.2:2023-01-20 10:26:00,008+1100 DEBUG [] Caesium-1-1 anonymous    breaklotsofstuff MailFetcherService run() method has been called
atlassian-jira-incoming-mail.log.1:2023-01-20 10:28:07,723+1100 DEBUG [] Caesium-1-2 anonymous    breaklotsofstuff MailFetcherService run() method has been called

      In the example below, the full reindex ran for more than 1 minute, delaying the MailFetcherService until the reindex completed.

      2023-01-20 10:26:26,681+1100 JiraTaskExecutionThread-7 INFO admin 626x1730x1 rosp4s 103.217.167.153,0:0:0:0:0:0:0:1 /secure/admin/IndexReIndex!reindex.jspa [c.a.j.w.a.admin.index.IndexAdminImpl] Re-indexing started
...
2023-01-20 10:28:07,919+1100 JiraTaskExecutionThread-7 INFO admin 626x1730x1 rosp4s 103.217.167.153,0:0:0:0:0:0:0:1 /secure/admin/IndexReIndex!reindex.jspa [c.a.j.w.a.admin.index.IndexAdminImpl] Re-indexing is 100% complete. Current index: 
2023-01-20 10:28:07,919+1100 JiraTaskExecutionThread-7 INFO admin 626x1730x1 rosp4s 103.217.167.153,0:0:0:0:0:0:0:1 /secure/admin/IndexReIndex!reindex.jspa [c.a.j.util.index.CompositeIndexLifecycleManager] Reindex All COMPLETED without any errors. Total time: 100991ms. Reindex run: 4
2023-01-20 10:28:07,919+1100 JiraTaskExecutionThread-7 INFO admin 626x1730x1 rosp4s 103.217.167.153,0:0:0:0:0:0:0:1 /secure/admin/IndexReIndex!reindex.jspa [c.a.j.w.a.admin.index.IndexAdminImpl] Re-indexing finished

      Workaround

      Currently there is no known workaround for this behavior. A workaround will be added here when available

            Assignee:
            Unassigned
            Reporter:
            Konde
            Votes:
            1 Vote for this issue
            Watchers:
            6 Start watching this issue

              Created:
              Updated: