Uploaded image for project: 'Jira Data Center'
  1. Jira Data Center
  2. JRASERVER-45340

Failing to generate a thumbnail should remember that the operation failed

XMLWordPrintable

      NOTE: This bug report is for JIRA Server. Using JIRA Cloud? See the corresponding bug report.

      Summary

      If an exception occurs during the processing of a thumbnail (such as a corrupt jpg), JIRA will not record the failure, and the next time the resource is requested, it will generate it (and fail) again.

      Typically on those issue that failed due to the condition:

      2015-09-11 21:24:41.272460500 2015-09-11 21:24:41,267 ajp-nio-127.0.0.104-8009-exec-1063 WARN sysadmin 1283x137820x1 t77wfs 202.68.74.170,104.192.141.33 /browse/DD-108 [c.a.j.issue.thumbnail.DefaultThumbnailManager] Error writing to thumbnail file: /data/jirastudio/jira/home/data/attachments/TEST/10000/TEST-108/thumbs/_thumb_45662.png1391332809470942590.tmp
2015-09-11 21:24:41.272467500 com.atlassian.jira.issue.attachment.AttachmentReadException: Remote blobstore couldn't provide an input stream for attachment 45662: java.awt.image.RasterFormatException: (y + height) is outside of Raster
2015-09-11 21:24:41.272795500 	at sun.awt.image.ByteInterleavedRaster.createWritableChild(ByteInterleavedRaster.java:1248)
2015-09-11 21:24:41.272796500 	at sun.awt.image.ByteInterleavedRaster.createChild(ByteInterleavedRaster.java:1208)
2015-09-11 21:24:41.272797500 	at com.twelvemonkeys.imageio.plugins.jpeg.JPEGImageReader.readImageAsRasterAndReplaceColorProfile(Unknown Source)
2015-09-11 21:24:41.272800500 	at com.twelvemonkeys.imageio.plugins.jpeg.JPEGImageReader.read(Unknown Source)
2015-09-11 21:24:41.272800500 	at com.atlassian.core.util.thumbnail.Thumber.scaleImage(Thumber.java:231)
2015-09-11 21:24:41.272800500 	at com.atlassian.core.util.thumbnail.Thumber.createThumbnail(Thumber.java:291)
2015-09-11 21:24:41.272812500 	at com.atlassian.core.util.thumbnail.Thumber.retrieveOrCreateThumbNail(Thumber.java:270)

      Expected Result

      This should be tracked even in the event of catastrophic failure (such as OOME taking down the application), and thus the attempt should be recorded and committed before the attempt, not after.

      When it fails, it should disable the thumbnail from being generated again by marking thumbnailable on the fileattachment table to 0

      Actual Result

      It will take a very long time to load the issue, and fail the page load. Multiple reload will still hit into the same issue. Potential piling up the memory if it keeps loaded.

      Workaround

      Identify from the error message which image is causing the error, for example in the above error it is image id 45662. Shut down JIRA, take a backup of the database and then run:

      update fileattachment set thumbnailable = 0 where id in (<image ids you identified above>)

            [JRASERVER-45340] Failing to generate a thumbnail should remember that the operation failed

            Nick Clarke added a comment -

            I merged this to master today: https://stash.atlassian.com/projects/JIRA/repos/jira/pull-requests/5619/overview
            Some other changes went from 7_0 to master shortly before: https://stash.atlassian.com/projects/JIRA/repos/jira/commits/2e4a6c973912e350200d0e62accb13702c9e3ee2

            Nick Clarke added a comment - I merged this to master today: https://stash.atlassian.com/projects/JIRA/repos/jira/pull-requests/5619/overview Some other changes went from 7_0 to master shortly before: https://stash.atlassian.com/projects/JIRA/repos/jira/commits/2e4a6c973912e350200d0e62accb13702c9e3ee2

            Oswaldo Hernandez (Inactive) added a comment -

            Update

            Hi all,

            Unfortunately, only a part of this fix was shipped in 7.0.4 / 7.1.0-OD-02, which was the change to jira-core such that JIRA won't try forever to render an unrenderable thumnail. Once JIRA tries for the first time, it will record it in the database and not try again.

            The part that did not make it in was the upgrade to our image-io implementation (12 monkeys library) that fixed the issue where images of certain (very specific) proportions caused it to generate broken thumbnails.

            This upgrade had to be done in a branch of atlassian-core and unfortunately, there was a hiccup in the process and we missed upgrading atlassian-core in jira to the version containing the 12 monkeys library upgrade.

            We will be doing this upgrade now and it should be out in 7.0.10 and 7.1.0-OD-06.

            Regards,

            Oswaldo Hernández.
            JIRA Bugmaster.
            [Atlassian].

            Oswaldo Hernandez (Inactive) added a comment - Update Hi all, Unfortunately, only a part of this fix was shipped in 7.0.4 / 7.1.0-OD-02, which was the change to jira-core such that JIRA won't try forever to render an unrenderable thumnail. Once JIRA tries for the first time, it will record it in the database and not try again. The part that did not make it in was the upgrade to our image-io implementation (12 monkeys library) that fixed the issue where images of certain (very specific) proportions caused it to generate broken thumbnails. This upgrade had to be done in a branch of atlassian-core and unfortunately, there was a hiccup in the process and we missed upgrading atlassian-core in jira to the version containing the 12 monkeys library upgrade. We will be doing this upgrade now and it should be out in 7.0.10 and 7.1.0-OD-06. Regards, Oswaldo Hernández. JIRA Bugmaster. [Atlassian] .

            Mateus Bueno added a comment -

            It's critical. The JIRA goes down a lot of times.

            Mateus Bueno added a comment - It's critical. The JIRA goes down a lot of times.

            Oswaldo Hernandez (Inactive) added a comment -

            Uhmmm, not really sure about the prescribed solution here... What about actually fixing the broken original image as a solution to the issue instead of adding workarounds like this one?

            I do believe our thumbnailing code could be hardened to tolarate failures during IO better, but we need specifics to make concrete changes.

            Cheers,
            Os.

            Oswaldo Hernandez (Inactive) added a comment - Uhmmm, not really sure about the prescribed solution here... What about actually fixing the broken original image as a solution to the issue instead of adding workarounds like this one? I do believe our thumbnailing code could be hardened to tolarate failures during IO better, but we need specifics to make concrete changes. Cheers, Os.

              ohernandez@atlassian.com Oswaldo Hernandez (Inactive)
              hmurn metrics
              Affected customers:
              2 This affects my team
              Watchers:
              20 Start watching this issue

                Created:
                Updated:
                Resolved: