Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-138238Backlog
Update hooks that run resharding code paths to expect concurrent reshardings
currently the only hook that starts resharding code paths is {{add_remove_shards.py}} and it is included by various suites to run as a background hook. that test hook itself expects/tolerates ReshardCollectionInProgress , and the various tests in the suites that run alongside it also tolerate ReshardCollectionInProgress and ConflictingOperationInProgress (some locally, others via shell overrides).
This ticket should
# relax the toleration and update the handled errors to be the capped resha
SERVER-138298Backlog
Pull out max reshardings cap into a cluster parameter and restore default max thread counts
Add a new cluster parameter for the maximum number of concurrent resharding operations that are allowed.
{{Write to sharding_cluster_parameters.idl}} which is for cluster-topology state such as shard cardinality and add/remove-shard in progress. We want all shards to agree on the cap (wouldn't be safe for one shard to think maximum is N and another shard to think the maximum resharding operations allowed is M).
The default value of the cap should be 10.
This ticket should land _after_ S
SERVER-138234In Progress
Cleanup legacy code branches for resharding refresh path and FCV pinning
Now that v9.0 is last LTS, there is a lot of code branching across resharding coordinator and participants for legacy refresh path and FCV pinning in opCtx. We should get rid of this code including any tests.
SERVER-138370Needs Scheduling
Add namespace to shard-wide resharding failpoints
Many tests (~35) that use the ReshardingTest fixture use failpoints that pause or error on every resharding operation on a a shard (eg. reshardingPause*, pauseAfterRecipientReceivedCloneCmd, hangAfterInitializingIndexBuild). A namespace filtering should be added on the server side for these and the tests that use them should be updated to pass in the namespace (and drop the opt-out tags if they had been placed by SERVER-138368). Go in order of coverage percentage. then background concurrent resh
SERVER-137688In Progress
Define the pre and post conditions of the Resharding Coordinator state machine
SERVER-135224In Progress
Allow setting resharding parameters per operation
Today, resharding's configuration is based on a [number of server parameters|https://github.com/10gen/mongo/blob/f9aa94518f6dacfee609c921e53430823a8783f0/src/mongo/db/s/resharding/resharding_server_parameters.idl]. This causes friction in a few ways:
* Users must set these parameters on each mongod where resharding will be running (including secondaries, in the case of failovers) for them to take affect.
* Atlas users are not permissioned to change server parameters, so modifying these requi
SERVER-134813Open
Retune and stabilize reshard_collection_burst_locust
Check BF-46227 for analysis comments.
This issue is easily reproduceable at standard load. After running a sweep of various load levels, a 10% reduction in read/write load stabilizes the workload and produces consistent results, while still maintaining significant lag during resharding.
See the PR for graphs and metrics.
SERVER-138247Needs Scheduling
Create tests in the jstestfuzzer that run concurrent reshardings
SERVER-126091Open
Complete TODO listed in SERVER-106990
There is a TODO in the codebase referencing a resolved ticket which is assigned to you.
Please follow this link to see the lines of code referencing this resolved ticket:
https://github.com/10gen/mongo/search?q=SERVER-106990&type=Code
The next steps for this ticket are to either remove the outdated TODO or follow the steps in the TODO if it is correct. If the latter, please update the summary and description of this ticket to represent the work you're actually doing.
SERVER-111230Backlog
reshardCollection may fail to clean up metadata on its temporary collection when aborted
During the development of SERVER-111054, the team discovered that a {{reshardCollection}} operation that gets aborted while the donor is entering its {{donating-initial-data}} phase may fail to clean the metadata about its temporary collection from the global catalog, leading to {{checkMetadataConsistency}} potentially raising spurious alerts.
Affected Versions
8.3.0-rc0
SERVER-134413In Progress
Skip range deletions when transitioning from embedded to dedicated config servers
Original description from [SERVER-103990|https://jira.mongodb.org/browse/SERVER-103990]:
"Today, when a cluster transitions from embedded to dedicated config servers we use transitionToDedicatedConfigServer() which uses removeShard().
We drain sharded collections and then use range deletions to remove any data from the embedded config server safely in case it gets added back to the cluster as a config shard. For any other removeShard operation, we don't complete range deletions because th
SERVER-117751Open
Decouple balancing of config.system.sessions from sh.stopBalancer()
The ask in this ticket is to always balance config.system.sessions even when balancer is stopped.
Customer problem:
When the balancer is stopped, the config.system.sessions collection is not balanced, which can lead to significant operational issues. The most critical impact is that all session-related workload (such as session creation, updates, and TTL deletions) is concentrated on a single shard, rather than being distributed across the cluster. This can cause:
Increased write latency a
SERVER-125892Blocked
Add HostUnreachable to retryable errors in implicitly_retry_on_config_stepdowns.js override
Root Cause:
SERVER-108482 correctly removed the isNotPrimaryError retry from kNotIdempotent in Shard::remoteIsRetriableError, which had previously masked config-stepdown errors by silently retrying setDefaultRWConcern at the C++ layer before they reached the JS test layer. With that safety net gone, when the config primary steps down during write-concern satisfaction of setDefaultRWConcern, the response surfaces as writeConcernError: "code: 6, HostUnreachable, "operation was interrupted" ": an
SERVER-123150Backlog
Add optimization for unshardCollection when collection only has chunk(s) on destination shard
When unshardCollection is called on a sharded collection that only has chunk(s) residing on the destination shard, the full resharding operation cloning mechanism is unnecessary.
Add a fast path that skips data movement entirely in that case and only updates the collection metadata.
SERVER-136491Backlog
Remove Feature Flag for SPM-4934
Feature flags should be proactively removed to address technical debt.
See the [Feature Flag Github page|https://github.com/mongodb/mongo/blob/master/src/mongo/db/repl/FCV_AND_FEATURE_FLAG_README.md#feature-flags] for additional guidance.
SERVER-133485In Progress
Improve how we calculate recipient fetch delay
We use [now - postBatchResumeTokenTs|https://github.com/10gen/mongo/blob/419942d6ceda18113c25e5995c1b9e0e42c279f5/src/mongo/db/s/resharding/resharding_oplog_fetcher.cpp#L231]. The issue is that ts only has second granularity and we use the full ms value of now. This means that if the theoretical time for the resume token was 08:30.940 and time now is 80:30:950, the fetch delay will be reported as 950ms (950 - 0, 0 since ts has second granularity so it is rounded down) instead of the actual 10ms.
SERVER-135283In Progress
Add more observability improvements for resharding txn cloner/apply
We should add metrics to show progress: lsid cloned and applied. It will also be useful to add some indication that we are done with txn cloning and is now doing txn apply and how far in the oplog it has caught up.
SERVER-137733In Progress
Pre-conditions/Post-condition verification for the Resharding Recipient