[MDEV-17955] RocksDB: Server Collapse with certain Queries Created: 2018-12-10 Updated: 2023-04-27 |
|
| Status: | Open |
| Project: | MariaDB Server |
| Component/s: | Storage Engine - RocksDB |
| Affects Version/s: | 10.3.11 |
| Fix Version/s: | 10.4 |
| Type: | Bug | Priority: | Major |
| Reporter: | Juan Telleria | Assignee: | Sergei Petrunia |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | index, rocksdb | ||
| Environment: |
Windows |
||
| Attachments: |
|
| Description |
|
When doing Bulk Loads, RocksDB gets stuck with some scheduled bulk load events with "Creating sort index". Under normal circumstances, the procedure should execute in a couple of minutes, but at time of reporting, it has been running for nearly 6 hours without any result...
And if I try to kill the query, it does not get killed, causing performance issues to the whole server, and forcing a serve shutdown. Which could be the root cause? |
| Comments |
| Comment by Sergei Petrunia [ 2018-12-11 ] | ||||
|
It is not clear why it got stuck. What does SELECT * FROM INFORMATION_SCHEMA.processlist show about the hanging query? Are there any other queries on the server that would be inserting rows into RocksDB tables? If not, can you check
Is it increasing? Also, can you check
. | ||||
| Comment by Juan Telleria [ 2018-12-13 ] | ||||
|
I cannot execute the previous check queries, as I already rebooted the server (And did migration back to InnoDB). It showed "Creating sort index" for a query which shall have executed in a few minutes, for more than 6 hours... And after trying to KILL it, I saw killed, but the connection and the query did not disappear. | ||||
| Comment by Juan Telleria [ 2018-12-13 ] | ||||
|
Maybe it is related to the table isolation level? No idea It also happened before however, but I could not see the backtrace for having no error log |