Skip to content
This repository has been archived by the owner on Jan 25, 2022. It is now read-only.

Don't lose service instances after gateway/broker restart #15

Merged

Conversation

rkoster
Copy link

@rkoster rkoster commented Sep 23, 2013

After a gateway/broker restart it is unable to rebuild its catalog from the cloud controller.
As a consequence you will not be able to bind to existing services.
Services which where bind before the restart keep functioning as normal.

The problem could be traced back to the addition of a type field for service instances.
This type field is not expected by the catalog manager and as a result it decides to skip the instance.

mark-rushakoff added a commit that referenced this pull request Sep 23, 2013
Don't lose service instances after gateway/broker restart
@mark-rushakoff mark-rushakoff merged commit d785db0 into cloudfoundry-attic:master Sep 23, 2013
@mark-rushakoff
Copy link
Contributor

Merged. We ran the tests for this on our internal Jenkins server and everything passed.

@mark-rushakoff / @matthewmcnew

cf-services pushed a commit that referenced this pull request Sep 23, 2013
@matt-royal
Copy link
Contributor

Newer versions of vcap_common ignore unknown fields in messages, which solves this problem without a change to vcap-services-base. It looks like vcap_services_base was just behind in its version requirements for vcap_common, so I pushed dc488d8, which bumps the required version of vcap_common. Commit a99b907 reverts this merge, since it is no longer necessary.

@rkoster rkoster deleted the fix-broken-catalog-update branch September 24, 2013 09:54
@rkoster
Copy link
Author

rkoster commented Sep 24, 2013

@matt-royal could you please reference the code which is responsible for ignoring the unknown fields? Because the underlining problem is still present in cf-services-release which already uses vcap-common 2.2.1.

@matt-royal
Copy link
Contributor

Here's the change to vcap-common: cloudfoundry/vcap-common@41db379. However, when we investigated further we found this line of code in ServiceMessage, the immediate parent of ServiceInstanceHandleV2: https://github.com/cloudfoundry/vcap-services-base/blob/master/lib/base/api/message.rb#L11. This line maintains the behavior we'd intended to remove. We plan to modify ServiceMessage as well, so that it also ignores unknown fields, which will solve your problem, and then will update the Gemfile for the mysql gateway.

We'll update this pull request when that's done.

@ryantang / @matt-royal

@matt-royal
Copy link
Contributor

We have this issue resolved, but are blocked from releasing it until we resolve a problem with our CI environment. This fix should be out early next week, if not sooner.

@matt-royal
Copy link
Contributor

We believe that commit 0b0ee2e in vcap-services-base should resolve your problem. We've also bumped the version of vcap-services-base used by the mysql service in this commit: cloudfoundry-community-attic/cf-services-release@3c8f492

Please let us know if the problem persists.

@rkoster
Copy link
Author

rkoster commented Oct 1, 2013

Thanks binding and unbinding services now works again after a gateway/broker restart.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants