XML_DEFINITION_DATA format may cause issues in Bamboo

[Brian Andersen]

We have had issues multiple times in our Bamboo instance where the TFS plugin causes errors connecting to repositories. Atlassian has done a root cause analysis and determined that the TFS plugin is storing data in a manner that is not default (old format); specifically the XML_DEFINITION_DATA column.

Please look into updating the plugin to be compliant with the new format of storing this information? Specific examples can be provided.

[Sergey Podobry]

Hi Brian,

Could you provide some background? (What is your Bamboo version, what errors do you encounter, steps to reproduce, and so on).

Note that the TFS plugin doesn't create any XML data, it uses only Atlassian provided interfaces. So it's strange to hear Atlassian blaming someone about XML data format.

[Brian Andersen]

Sergey:

We had this happen with Bamboo 5.10.x and 5.15.7 with both 1.17 and 1.18. of the plugin. When it happens, you get an Internal Server/500 error to all repository types, TFS and Bitbucket. I have attached the error which was thrown attempting to access a Bitbucket repo. It's not reproducible, but it has happened to us 3 times in the past year.

As for "blaming" Stellarity, I don't think that's the case. I suspect from what they said vs. your response, I suspect you may be using an obsoleted interface to write the XML data? They did specifically say it's the "old style" of storing this information.

[Brian Andersen]

Internal server error details

[Brian Andersen]

Also, please note I only opened this to make you aware for research. While it has impacted us, it's not reproducible nor is it frequent. There may be nothing you can or want to do, but opening an enhancement was just the next step.

[Sergey Podobry]

I see. Indeed there is a new API starting from Bamboo 5.15 (but you've said it happened to Bamboo 5.10.x, so I'm still confused).

We're working on migrating to the new API. I hope it will fix this issue.

Thanks for reporting!

[Brian Andersen]

No problem.

We didn't do a deep-dive with Atlassian last time it happened on 5.10.x, but it was something that required manual database clean-up. We/they may have just correlated the instances together incorrectly. At this point, our goal was just to relay the information about the API to you less so than try to figure out what was happening in the past, and it seems you are on it.

Thanks!