[Test] Allow allocation in mixed cluster #129680
Merged
Conversation
This file contains hidden or 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
The RunningSnapshotIT upgrade test adds shutdown marker to all nodes and removed them once all nodes are upgraded. If an index gets created in a mixed cluster, for example by ILM or deprecation messages, the index cannot be allocated because all nodes are shutting down. Since the cluster ready check between node upgrades expects a yellow cluster, the unassigned index prevents the ready check to succeed and eventually timeout. This PR fixes it by removing shutdown marker for the 1st upgrade node to allow it hosting new indices. Resolves: elastic#129644 Resolves: elastic#129645 Resolves: elastic#129646
ywangd
added
>test
elasticsearchmachine
added
the
Team:Distributed Coordination
Collaborator
|
Pinging @elastic/es-distributed-coordination (Team:Distributed Coordination) |
nicktindall
approved these changes
Jun 19, 2025
Contributor
nicktindall
left a comment
nicktindall
left a comment
There was a problem hiding this comment.
Nice, so I assume this still doesn't allow the snapshot to complete during the upgrade because there will be 2 shards that can't be assigned, because only 1 node has no shutdown marker?
Member
Author
Those two shards remain on their initial nodes. They are not unassigned because they are not new shards. The snapshot cannot complete because:
So yeah the snapshot can completely only when all nodes are upgraded and remaining shutdown marker removed. |
ywangd
added
the
auto-merge-without-approval
Member
Author
|
@elasticmachine update branch |
kderusso
pushed a commit
to kderusso/elasticsearch
that referenced
this pull request
Jun 23, 2025
The RunningSnapshotIT upgrade test adds shutdown markers to all nodes and removes them once all nodes are upgraded. If an index gets created in a mixed cluster, for example by ILM or deprecation messages, the index cannot be allocated because all nodes are shutting down. Since the cluster ready check between node upgrades expects a yellow cluster, the unassigned index prevents the ready check to succeed and eventually timeout. This PR fixes it by removing shutdown marker for the 1st upgrade node to allow it hosting new indices. Resolves: elastic#129644 Resolves: elastic#129645 Resolves: elastic#129646
mridula-s109
pushed a commit
to mridula-s109/elasticsearch
that referenced
this pull request
Jun 25, 2025
The RunningSnapshotIT upgrade test adds shutdown markers to all nodes and removes them once all nodes are upgraded. If an index gets created in a mixed cluster, for example by ILM or deprecation messages, the index cannot be allocated because all nodes are shutting down. Since the cluster ready check between node upgrades expects a yellow cluster, the unassigned index prevents the ready check to succeed and eventually timeout. This PR fixes it by removing shutdown marker for the 1st upgrade node to allow it hosting new indices. Resolves: elastic#129644 Resolves: elastic#129645 Resolves: elastic#129646
ywangd
added a commit
to ywangd/elasticsearch
that referenced
this pull request
Jul 30, 2025
This is the same failure as observed in elastic#129644 for which the original fix elastic#129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: elastic#132135 Resolves: elastic#132136 Resolves: elastic#132137
elasticsearchmachine
pushed a commit
that referenced
this pull request
Jul 31, 2025
This is the same failure as observed in #129644 for which the original fix #129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: #132135 Resolves: #132136 Resolves: #132137
ywangd
added a commit
to ywangd/elasticsearch
that referenced
this pull request
Jul 31, 2025
…2157) This is the same failure as observed in elastic#129644 for which the original fix elastic#129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: elastic#132135 Resolves: elastic#132136 Resolves: elastic#132137 (cherry picked from commit f39ccb5) # Conflicts: # muted-tests.yml
elasticsearchmachine
pushed a commit
that referenced
this pull request
Jul 31, 2025
…132233) This is the same failure as observed in #129644 for which the original fix #129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: #132135 Resolves: #132136 Resolves: #132137 (cherry picked from commit f39ccb5) # Conflicts: # muted-tests.yml
afoucret
pushed a commit
to afoucret/elasticsearch
that referenced
this pull request
Jul 31, 2025
…2157) This is the same failure as observed in elastic#129644 for which the original fix elastic#129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: elastic#132135 Resolves: elastic#132136 Resolves: elastic#132137
smalyshev
pushed a commit
to smalyshev/elasticsearch
that referenced
this pull request
Jul 31, 2025
…2157) This is the same failure as observed in elastic#129644 for which the original fix elastic#129680 did not really work. It did not work because the the ordering of checks. The shutdown marker is removed after the cluster passes ready check so that new shards can be allocated. But the cluster cannot pass the ready check before the shards are allocated. Hence the circular dependency. In hindsight, there is no need to put shutdown record for all nodes. It is only needed on the node that upgrades the last to prevent snapshot from completion during the upgrade process. This PR does that which ensures there are always 2 nodes for hosting new shards. Resolves: elastic#132135 Resolves: elastic#132136 Resolves: elastic#132137
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
The RunningSnapshotIT upgrade test adds shutdown markers to all nodes and removes them once all nodes are upgraded. If an index gets created in a mixed cluster, for example by ILM or deprecation messages, the index cannot be allocated because all nodes are shutting down. Since the cluster ready check between node upgrades expects a yellow cluster, the unassigned index prevents the ready check to succeed and eventually timeout. This PR fixes it by removing shutdown marker for the 1st upgrade node to allow it hosting new indices.
Resolves: #129644
Resolves: #129645
Resolves: #129646