-
Notifications
You must be signed in to change notification settings - Fork 577
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Partition autobalancer full disk test #5839
Merged
ztlpn
merged 8 commits into
redpanda-data:dev
from
ztlpn:partition-autobalancer-test-full-disk
Aug 5, 2022
Merged
Partition autobalancer full disk test #5839
ztlpn
merged 8 commits into
redpanda-data:dev
from
ztlpn:partition-autobalancer-test-full-disk
Aug 5, 2022
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ztlpn
requested review from
dotnwat,
NyaliaLui,
mmaslankaprv and
VadimPlh
as code owners
August 4, 2022 13:54
ztlpn
force-pushed
the
partition-autobalancer-test-full-disk
branch
from
August 4, 2022 13:56
55b1d1e
to
957d994
Compare
mmaslankaprv
reviewed
Aug 4, 2022
mmaslankaprv
reviewed
Aug 4, 2022
mmaslankaprv
reviewed
Aug 4, 2022
ci failure is #5713 |
ztlpn
force-pushed
the
partition-autobalancer-test-full-disk
branch
from
August 5, 2022 00:28
957d994
to
55dfad0
Compare
mmaslankaprv
previously approved these changes
Aug 5, 2022
When moving partitions away from unavailable nodes it is desireable to violate disk ratio that we use to determine when we need to move nodes away from a node - this will allow us to save data if we have some free space left. But we shouldn't go to 100%, so we use the storage_space_alert_free_threshold_percent config value to determine the hard limit.
Previously there was a following bug with moving partitions away from nodes with full disks: because the planner tries to move all replicas in a set from "bad" nodes in one reallocation, if we don't take into account previous reallocations, we might consider that some node is still full even if we have already planned several moves away from it. This lead to excessive movements being planned: when a full node became almost empty after executing a batch of reallocations. In this commit we use "final" node disk usage (after all reallocations are finished) where appropriate.
ztlpn
force-pushed
the
partition-autobalancer-test-full-disk
branch
from
August 5, 2022 09:48
55dfad0
to
a484b37
Compare
rebased on dev to resolve a merge conflict |
mmaslankaprv
approved these changes
Aug 5, 2022
ci failure in the release build is #5276 and in the debug build is #5713 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Cover letter
Add a test for partition balancer full disk node handling with the following scenario:
To make this test possible, we introduce an environment variable that forces redpanda to report mock disk size in the health monitor.
This test uncovered a bug in the partition balancer planner: because the planner tries to move all replicas in a set from "bad" nodes in one reallocation, if we don't take into account previous reallocations, we might think that some node is still full even if we have already planned several moves away from it. This lead to excessive movements being planned: when a full node became almost empty after executing a batch of reallocations. To fix this we use "final" node disk usage (after all reallocations are finished) where appropriate.
Backport Required
UX changes
none
Release notes