|
Ran
|
Files
518
|
Run time
23s
|
Badge
README BADGES
|
push
github
test: fix flakiness of the election + qsync stress test The source of flakiness of the test came from its sensitivity to the replication timeout, and, consequently, the leader death timeout: the leader could get re-elected while we were trying to send an insert request to the old leader — we would never notice the ER_READONLY error. To fix this, let's increase the replication timeout to 1 second (the default value), which seems to give enough stability. While we are here, port the test to luatest to simplify its maintenance with the following principal changes: 1. After a new leader gets elected, check the emptiness of his limbo rather than writing an asynchronous transaction. 2. After the old leader gets restarted, wait for cluster synchronization with him. Closes #10543 NO_CHANGELOG=<flaky test> NO_DOC=<flaky test> (cherry picked from commit d97ac05b8)
70043 of 123907 branches covered (56.53%)
103067 of 117807 relevant lines covered (87.49%)
2129582.96 hits per line
| Coverage | ∆ | File | Lines | Relevant | Covered | Missed | Hits/Line | Branch Hits | Branch Misses |
|---|