If the microsummary service attempts to update a microsummary, and the update attempt fails, the service tries again in 15 seconds, which means that if the failure occurred during or after the generator/page was requested, and the failure is persistent, then the service will repeatedly issue requests every 15 seconds, possibly inadvertently DOSing the server and exacerbating any load/availability problems it is experiencing. The microsummary service should never attempt to update a microsummary more frequently than the default interval. This should be easy to fix. We just need to make the service reset the microsummary's expiration property before it starts the update instead of after the update completes. But making this depend on bug 339543, as this probably involves touching code that bug is changing.
Created attachment 228121 [details] [diff] [review] patch v1: the minimal fix for this bug Here's the minimal fix for this bug. It updates the expiration time before trying to refresh the microsummary during the regular updateMicrosummaries cycle. But it doesn't yet remove the code that updates the expiration time after the microsummary is refreshed. We should move that code so that it only runs when microsummaries are manually refreshed (i.e. not during the regular update cycle, but because of some user action), but perhaps we can do that in a separate patch.
Comment on attachment 228121 [details] [diff] [review] patch v1: the minimal fix for this bug email@example.com
Fix checked in to trunk.
Comment on attachment 228121 [details] [diff] [review] patch v1: the minimal fix for this bug a=darin on behalf of drivers