Improve the reliability of the kills from fdbcli #2512
Conversation
…prove the reliability of the kill process secondaryAddresses are included in the list of processes which can be killed
|
This closes #2209 |
| Key full_ip_port2 = StringRef(workerInterf.reboot.getEndpoint().addresses.secondaryAddress.get().toString()); | ||
| StringRef ip_port2 = full_ip_port2.endsWith(LiteralStringRef(":tls")) ? full_ip_port2.removeSuffix(LiteralStringRef(":tls")) : full_ip_port2; | ||
| (*address_interface)[ip_port2] = std::make_pair(kv.value, leaderInterf); | ||
| } |
There was a problem hiding this comment.
This bit is new, is it necessary to issue kills to both interfaces? Is this going to print duplicates in our list?
There was a problem hiding this comment.
It shouldn't be necessary to issue kills to both interfaces, but we want to allow people to issue kills to either interface. I've found that it's hard to predict which interface is going to appear in the kill list today.
There was a problem hiding this comment.
If we happen to issue kills to the same process twice because of this, is it possible that the process gets killed twice, or is there something that protects us from that?
There was a problem hiding this comment.
Once a worker is killed, it will come back with a different interface, so it is safe to send multiple kills.
Co-Authored-By: A.J. Beamon <[email protected]>
3 participants
No description provided.