Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.6.0, 7.1.4, 7.1.0, 7.1.1, 7.1.2, 7.1.3, 7.1.5
-
0
-
Yes
Description
If PrepareTopologyChange takes more than 60 seconds the rebalance fails.
Currently host name resolution (strictly mapping of UUID to host with IP lookup) takes place during the call, in a serial fashion. If resolution is slow the call may exceed the time limit.
Workarounds:
- Add/remove one Query node at a time (to avoid compounding any delay)
- Ensure host name resolution is fast
- Perhaps even temporarily listing the names in local hosts files and ensuring they're checked first (before DNS or LDAP)
Fixes:
- Perform the host name resolution in parallel.
- Perform the host name resolution only for and when ejecting nodes.
Attachments
Issue Links
- is a backport of
-
MB-58732 Topology change: PrepareTopologyChange exceeding call time limit
- Closed
For Gerrit Dashboard: MB-59546 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
199558,4 | MB-59546 Perform preparations in parallel | 7.2.3 | query | Status: MERGED | +2 | +1 |
200381,2 | MB-59528 MB-59530 MB-59531 MB-59532 MB-59533 MB-59546: 7.1.6-MP1 | 7.1.x | query | Status: MERGED | +2 | +1 |