Commits on Sep 6, 2024

1. k/group_manager: return not_coordinator quickly in tx operations …

   group_manager::attached_partition::catchup_lock can get blocked for
   extended periods of time. For example in the following scenario:
   1. consumer_offsets partition leader gets isolated
   2. some group operation acquires a read lock and tries to replicate a
     batch to the consumer_offsets partition. This operation hangs for an
     indefinite period of time.
   3. the consumer_offsets leader steps down
   4. group state cleanup gets triggered, tries to acquire a write lock,
     hangs until (2) finishes
   
   Meanwhile, clients trying to perform any tx group operations will get a
   coordinator_load_in_progress errors and blindly retry, without even
   trying to find the real coordinator.
   
   Check for leadership without the read lock first to prevent that (this
   is basically a "double-check" pattern as we have to check the second
   time under the lock.)
   
   (cherry picked from commit 440ed2c)

   

   ztlpn committed Sep 6, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for cbd4cdc
   • Browse repository at this point

   Copy the full SHA

   cbd4cdc View commit details

   Browse the repository at this point in the history