Heavy loaded server not serving queries after start
- Dominant language
- Java
- Stars
- 6.1k
- Forks
- 1.5k
- Avg merge
- 1d 21h
- Merged PRs (30d)
- 189
Description
For a server with a lot of segments for Upsert tables, the server start process may result in high GC activity and its zk client may be disconnected temporarily. In such case, the server may not register itself to serve query traffic because the following zk config writes fails -- note that the other server start process still works fine. So a solution here is to at least retry the config write instead of giving it up after one attempt.
https://github.com/apache/pinot/blob/7e9ca6a5a4afe0d4e283ac1307c45430e474cbf2/pinot-server/src/main/java/org/apache/pinot/server/starter/helix/BaseServerStarter.java#L468
Contributor guide
Research direction
Start in pinot-server/src/main/java/org/apache/pinot/server/starter/helix/BaseServerStarter.java around line 468 and trace the ZooKeeper configuration write during server startup. Confirm the failure path when the client is temporarily disconnected; done means the configuration write is retried so the server can register to serve query traffic.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java
- Domain
- distributed-systems
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100