Uploaded image for project: 'translate5'
  1. translate5
  2. TRANSLATE-1654

TermTagger stops working on import of certain task

XMLWordPrintable

      This is really strange. A certain task led again and again to a crash of the termTagger. But after eventually after multiple times restarting the termTagger the task was successfully imported, I exported the task and it was not reproduceable with this package (TBX is part of the package). A possibility would be, that there are interferences between different tasks, that use the same TBX and run at the same time, because in the concrete situation there where 9 different import packages in the import and one package had been imported twice. And in all 10 imports the same TBX had been used. Languages had been DE and zh_CN.

       

          [TRANSLATE-1654] TermTagger stops working on import of certain task

          Thomas Lauria added a comment -

          marcmittag

          To answer your above questions, I will do the following changes:

          1. If on import no termtagger is available at all (E1131) I stop task import and put the task into state error (as discussed on call some days ago)
          2. For the not reachable for a few requests: I will find a usable solution, only timeout errors should be affected here
          3. No cron call: See TRANSLATE-1161 and TRANSLATE-392, and https://confluence.translate5.net/display/TIU/Install+translate5+and+direct+dependencies section Configure Cronjobs / task scheduler

          Thomas Lauria added a comment - marcmittag To answer your above questions, I will do the following changes: If on import no termtagger is available at all (E1131) I stop task import and put the task into state error (as discussed on call some days ago) For the not reachable for a few requests: I will find a usable solution, only timeout errors should be affected here No cron call: See TRANSLATE-1161 and TRANSLATE-392 , and https://confluence.translate5.net/display/TIU/Install+translate5+and+direct+dependencies section Configure Cronjobs / task scheduler

          Marc Mittag [Administrator] added a comment -

          tlauria What happens once in a while is, that a termtagger is only not reachable for a few requests. Probably because it has to much load. This must be looked for, when automatically disabling termtaggers. And what happens, if no cron call is configured? How will the termtaggers be reactivated?

          Marc Mittag [Administrator] added a comment - tlauria What happens once in a while is, that a termtagger is only not reachable for a few requests. Probably because it has to much load. This must be looked for, when automatically disabling termtaggers. And what happens, if no cron call is configured? How will the termtaggers be reactivated?

          Thomas Lauria added a comment -

          The concrete problem of crashing termtaggers could not be reproduced. The whole termtagger logging and internal error handling was refactored right now.

          Also an internal lock list of not available termtaggers was introduced. So a not reachable termtagger is not queried again and again with data to tag.

          The periodical cronjob checks for the availability of the termtaggers and cleans the above mentioned lock. So if the termtagger is up again, the lock can either manually removed from the Zf_memory table, or after at least 15 Minutes the periodical cron does the clean up.

          See also https://confluence.translate5.net/display/TAD/EventCodes#EventCodes-E1129.

          Thomas Lauria added a comment - The concrete problem of crashing termtaggers could not be reproduced. The whole termtagger logging and internal error handling was refactored right now. Also an internal lock list of not available termtaggers was introduced. So a not reachable termtagger is not queried again and again with data to tag. The periodical cronjob checks for the availability of the termtaggers and cleans the above mentioned lock. So if the termtagger is up again, the lock can either manually removed from the Zf_memory table, or after at least 15 Minutes the periodical cron does the clean up. See also https://confluence.translate5.net/display/TAD/EventCodes#EventCodes-E1129 .

          Thomas Lauria added a comment -

          Possibly related. Does our worker mechanism to ensure that only a specific amount of workers running at a time working properly?

          Thomas Lauria added a comment - Possibly related. Does our worker mechanism to ensure that only a specific amount of workers running at a time working properly?

            tlauria Thomas Lauria
            marcmittag Marc Mittag [Administrator]
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: