You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The regular tests do leak detection in test_helper.tcl, but the Sentinel and Cluster tests are managed by instances.tcl which has no leak detection.
I don't see an easy way to incorporate leak detection in instances.tcl, so this is a TODO for us to figure it out eventually. We may even want to consolidate how tests are run. Ideally we should have leak detection before kill_instance and as part of cleanup at the end (or, make a the_end for instances.tcl) too.
Also, the non-main tests (Sentinel, Cluster) sometimes leak processes after failures. Maybe we're missing some pid tracking or error catching so we can kill servers after test failures/crashes?
(The absence of leak detection on non-main tests probably allowed the recent Sentinel leak to go undetected in the wild for a long time.)
The text was updated successfully, but these errors were encountered:
The regular tests do leak detection in
test_helper.tcl
, but the Sentinel and Cluster tests are managed byinstances.tcl
which has no leak detection.I don't see an easy way to incorporate leak detection in
instances.tcl
, so this is a TODO for us to figure it out eventually. We may even want to consolidate how tests are run. Ideally we should have leak detection beforekill_instance
and as part ofcleanup
at the end (or, make athe_end
forinstances.tcl
) too.Also, the non-main tests (Sentinel, Cluster) sometimes leak processes after failures. Maybe we're missing some pid tracking or error catching so we can kill servers after test failures/crashes?
(The absence of leak detection on non-main tests probably allowed the recent Sentinel leak to go undetected in the wild for a long time.)
The text was updated successfully, but these errors were encountered: