Datasets:
Formats:
json
Languages:
English
Size:
< 1K
Tags:
distributed-systems
debugging
root-cause-analysis
fault-localization
agents
software-engineering
License:
instance_id stringlengths 8 20 | system stringclasses 11
values | issue_number stringlengths 4 6 | lite bool 2
classes | title stringlengths 0 133 | description stringlengths 150 30.7k | branch stringlengths 5 40 | issue_url stringlengths 41 54 | fix_url stringlengths 39 82 ⌀ | issue_created timestamp[s]date 2016-06-09 00:00:00 2026-04-22 00:00:00 ⌀ | severity stringclasses 3
values | root_cause_type stringclasses 7
values | bug_complexity stringclasses 5
values | distributed bool 2
classes | contains_analysis bool 2
classes | analysis_removed stringclasses 3
values | contains_logs bool 2
classes | contains_stack_traces bool 2
classes | num_artifacts int64 0 9 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
elasticsearch-59649 | elasticsearch | 59649 | false | Dedicated voting-only master-eligible node requires access to snapshot repositories for verification | Today a snapshot repository verification ensures that all master-eligible and data nodes have write access to the snapshot repository (and can see each other's data) since taking a snapshot requires data nodes and the currently-elected master to write to the repository. However, a dedicated voting-only master-eligible ... | 24143b07a47339e9455b5b2d0b20b5f6a352134e | https://github.com/elastic/elasticsearch/issues/59649 | https://github.com/elastic/elasticsearch/pull/61016 | 2020-07-15T00:00:00 | Major | Logic | Medium | true | false | No | false | false | 0 |
elasticsearch-85839 | elasticsearch | 85839 | false | RecoverySourceHandler#runWithGenericThreadPool caused deadlock | We saw a benchmark of a 7.17.2 cluster get stuck with all generic threads blocked in RecoverySourceHandler#runWithGenericThreadPool (see many-shards-threaddump.txt.gz):
$ python ~/src/jstack2json/jstack2json.py many-shards-threaddump.txt | jq '.[].threads[] | select(.elasticsearch.threadpool == "generic")' -cMr | wc -... | 44fa734bbd60a571dfa49d8e80440df9fc85e105 | https://github.com/elastic/elasticsearch/issues/85839 | https://github.com/elastic/elasticsearch/pull/86127 | 2022-04-12T00:00:00 | Major | Concurrency | Hard | true | false | N/A | false | true | 0 |
elasticsearch-87237 | elasticsearch | 87237 | false | StoreRecovery#restore may invoke Repository#restoreShard on transport thread | This code involves getRepositoryData which may complete its listener on a transport worker:
elasticsearch/server/src/main/java/org/elasticsearch/index/shard/StoreRecovery.java
Lines 528 to 547 in 898d849
final StepListener<IndexId> indexIdListener = new StepListener<>();
// If the index UUID was not found in the re... | 35b48729b183bc180914146a361609f01b08b921 | https://github.com/elastic/elasticsearch/issues/87237 | https://github.com/elastic/elasticsearch/pull/87264 | 2022-05-31T00:00:00 | Major | Concurrency | Medium | true | true | No | false | false | 0 |
elasticsearch-90442 | elasticsearch | 90442 | false | Disk indicator - Cover edge case where data node is out of space but has no indices. | Current behaviour
The disk indicator reports the following when a data node is yellow or red but has no indices.
Symptom:
1 node with roles: [data] is out of disk or running low on disk space.
Impact: None
Diagnosis: 0 indices reside on nodes that have run or are likely to run out of space, this can temporarily disab... | 9e6a784aa57877fcfe0352862356d4866aabb3ab | https://github.com/elastic/elasticsearch/issues/90442 | https://github.com/elastic/elasticsearch/pull/90772 | 2022-09-28T00:00:00 | Major | Error handling | Medium | false | true | No | false | false | 0 |
elasticsearch-93542 | elasticsearch | 93542 | false | Peer recovery may time out during recoverLocallyUpToGlobalCheckpoint | The recoverLocallyUpToGlobalCheckpoint step of peer recovery may take an extended period of time, but today we do not disable the recovery monitor during this step so if it takes more than 30 minutes (by default) then the recovery will time out, fail, and retry repeatedly. We should disable the recovery monitor during ... | 052a4f15b92713e422560ddee70050744f94555a | https://github.com/elastic/elasticsearch/issues/93542 | https://github.com/elastic/elasticsearch/pull/93551 | 2023-02-07T00:00:00 | Major | Performance | Hard | false | false | No | false | false | 0 |
elasticsearch-96544 | elasticsearch | 96544 | false | Cluster health computation sometimes runs on applier thread | In #90621 we moved most usages of the cluster health API onto a MANAGEMENT thread because it's too expensive to run on SAME in a large cluster. However, if the request includes any wait_for_* option then we set up a cluster state observer, which supplies its responses on the cluster applier thread. We need to fork thos... | 6caf9c1f60ba0b993dbb3d6cf97768275e9975fc | https://github.com/elastic/elasticsearch/issues/96544 | https://github.com/elastic/elasticsearch/pull/98904 | 2023-06-03T00:00:00 | Critical | Concurrency | Hard | true | false | No | false | false | 0 |
elasticsearch-98926 | elasticsearch | 98926 | false | Health API reports disk information Unknown status / No disk usage data symptom | Elasticsearch Service (ESS) users have observed the below error message when accessing the Health API of a deployment:\n\nError: Cannot read properties of undefined (reading 'replace\")\nat https://cloud.elastic.co/app.070493148d8f8558ab6.js:2:3103121\nat U1 (https://cloud.elastic.co/app.070493148d88558ab6.js:2:3103141... | cca26276b3965e24da9367c0ccfde411ec5124cd | https://github.com/elastic/elasticsearch/issues/98926 | https://github.com/elastic/elasticsearch/pull/105449 | 2023-08-28T00:00:00 | Major | Logic | Medium | false | true | No | true | true | 0 |
elasticsearch-105065 | elasticsearch | 105065 | false | Avoid race in LocalHealthMonitor w.r.t null values | There is a race condition in the LocalHealthMonitor regarding null values (i.e. initial/reset values):
Last reported value was null
there is new a health node
New run starts the update process and records previous value null
health node/master change
new reported value is null
Run sees null and considers that it's ok ... | bb1eddada3678257838b0590090ff9eb68acaa1b | https://github.com/elastic/elasticsearch/issues/105065 | https://github.com/elastic/elasticsearch/pull/105674 | 2024-02-02T00:00:00 | Major | Concurrency | Hard | true | true | No | false | false | 0 |
elasticsearch-110360 | elasticsearch | 110360 | false | PlainActionFuture#assertCompleteAllowed false positives on concurrent resolve | PlainActionFuture detects dangerous usage patterns where threads from the same pool are simultaneously resolving and waiting on the future. This is to prevent tests that deadlock and is only switched on when assertions are enabled.
This deadlock detection can falsely fire when threads from the same pool are simultaneo... | 89cd966b247148372aed73c3586b0754d6e176ca | https://github.com/elastic/elasticsearch/issues/110360 | https://github.com/elastic/elasticsearch/pull/110361 | 2024-07-02T00:00:00 | Major | Concurrency | Medium | true | true | No | false | false | 0 |
elasticsearch-113615 | elasticsearch | 113615 | false | Elasticsearch accepts script processors that are strings and not map | When creating an Ingest Pipeline, it's currently possible to create a processor, which is not a map:
PUT _ingest/pipeline/test
{
"processors": [
{
"script": """ctx.test=2;"""
}
]
}
This is accepted by Elasticsearch, however, the expected syntax should be:
PUT _ingest/pipeline/test
{
"processors": [... | b2a69b54e1f28711d423926d9261a8346fd22c76 | https://github.com/elastic/elasticsearch/issues/113615 | https://github.com/elastic/elasticsearch/pull/113697 | 2024-09-26T00:00:00 | Major | Error handling | Easy | false | false | N/A | false | true | 0 |
elasticsearch-116341 | elasticsearch | 116341 | false | Async search can throw NPE from onClusterResponseMinimizeRoundtrips | Stack trace seen from a customer environment:
`[2024-10-23T13:34:32,289][WARN ][o.e.a.s.SearchProgressListener] [data-2] [ibprod_azure] Failed to execute progress listener onResponseMinimizeRoundtrips
java.lang.NullPointerException: Cannot invoke "org.elasticsearch.xpack.search.MutableSearchResponse.updateResponseMini... | a4482d4c4c2d24418553510afa6a11a5f316983a | https://github.com/elastic/elasticsearch/issues/116341 | https://github.com/elastic/elasticsearch/pull/117504 | 2024-09-06T00:00:00 | Major | Concurrency | Medium | true | false | N/A | false | true | 0 |
elasticsearch-128583 | elasticsearch | 128583 | true | TransportIndicesStatsAction causes starvation of tasks on management thread pool due to Engine.getSparseVectorValueCount | Engine.getSparseVectorValueCount might block 100% of the management threads in case we're facing frequent cache misses.
This causes the thread pool to queue and potentially starve important other work that's also depending on the management thread pool.
This was discovered on an internal serverless project. | 041c42a77961b2a227e2631d2d4034d2716779c4 | https://github.com/elastic/elasticsearch/issues/128583 | https://github.com/elastic/elasticsearch/pull/128740 | 2025-05-28T00:00:00 | Major | Concurrency | Medium | false | false | N/A | true | false | 1 |
elasticsearch-128852 | elasticsearch | 128852 | true | ConcurrentModificationException in TransportClusterStatsAction | This error occurs:
java.util.ConcurrentModificationException
at java.base/java.util.ArrayList$Itr.checkForComodification(ArrayList.java:1096)
at java.base/java.util.ArrayList$Itr.next(ArrayList.java:1050)
at org.elasticsearch.server@9.1.0/org.elasticsearch.action.admin.cluster.stats.Ver... | c17bfcbfc219f1a0746baaf3b9501b2e479488a4 | https://github.com/elastic/elasticsearch/issues/128852 | https://github.com/elastic/elasticsearch/pull/130303 | 2025-06-03T00:00:00 | Major | Logic | Medium | true | false | N/A | false | true | 0 |
elasticsearch-134925 | elasticsearch | 134925 | false | Transforms will infinitely queue PIT closures causing a "PIT storm" | Transforms when using point-in-time (PIT) search will create multiple PITs over the lifetime of a single transform pivot page.
If there is any failure, or a new search is initiated, or the scheduled pivot completes, it will request the PITs to be deleted.
However, it will not wait for these PITs to be deleted before ... | 1904abbb730ec6bfc47a58425a3949664013ed55 | https://github.com/elastic/elasticsearch/issues/134925 | https://github.com/elastic/elasticsearch/pull/134955 | 2025-09-17T00:00:00 | Critical | Logic | Medium | true | true | No | false | false | 0 |
elasticsearch-135472 | elasticsearch | 135472 | true | GroupShardsIterator can significantly tie up transport workers | When querying many shards, it has been observed that iterating the shards themselves can end up tying up a transport_worker thread. I am not sure if we want to optimize this path further or pivot to a different thread when above some shard threshold.
Switching to another threadpool on few shards (<10k?) seems unhelpfu... | 8a38cbbafee64879a0fd20ce16e8546f30057cee | https://github.com/elastic/elasticsearch/issues/135472 | https://github.com/elastic/elasticsearch/pull/140747 | 2025-09-25T00:00:00 | Major | Performance | Medium | false | false | N/A | true | false | 0 |
elasticsearch-139252 | elasticsearch | 139252 | true | TransportGetTransformStatsAction can eat a transport thread, causing cluster instability | Transform stats does a fair bit of work just sitting in direct executor. This is not good. Depending on the scale of the cluster, if multiple users call the stats call for transforms, it could disconnect node from a cluster.
We should:
Make this way way cheaper, only consider transform tasks (do filtering directly in... | 8a3da5e87351f09306f75b74d29ce4ea1d80d28e | https://github.com/elastic/elasticsearch/issues/139252 | https://github.com/elastic/elasticsearch/pull/139857 | 2025-12-09T00:00:00 | Critical | Logic | Hard | true | true | No | true | false | 0 |
elasticsearch-140492 | elasticsearch | 140492 | false | NPE causing doc-level indexing failure in TSDB data stream | See e.g.:
Intake build failure
Local repro with extra logging
[2026-01-10T12:03:31,063][DEBUG][o.e.i.s.IndexingFailuresDebugListener] [test-cluster-0] index-fail [KCjEJ9R_BgO8TRX2QOd6dpR12oDh--qoyNZRQPy43y34Qdy2dpsyG0p___5kWKOUA7NNlnE] seq# [-2] allocation-id [[id=9ulrOz6xSemJQsht2vH7OQ]] primaryTerm [1] operationPrim... | 21fc8565454e1cea03fa7276dcb550f6a2bf6ec8 | https://github.com/elastic/elasticsearch/issues/140492 | https://github.com/elastic/elasticsearch/pull/140556 | 2026-01-10T00:00:00 | Major | Logic | Medium | false | true | No | true | true | 0 |
elasticsearch-141740 | elasticsearch | 141740 | false | Logging cancellation task with large number of nodes/actions | When canceling a possibly large list of actions / tasks, a possibly large amount of memory might be consumed at times as ES tries to log it by means of its description CancelTasksRequest#getDescription.
We should probably make sure the size of the underlying arrays doesn't become too big, and return a subset of the act... | 96be973fb5687c9837f51352fd7c8c5d2a5894aa | https://github.com/elastic/elasticsearch/issues/141740 | https://github.com/elastic/elasticsearch/pull/141815 | 2026-02-03T00:00:00 | Major | Performance | Medium | true | false | No | false | false | 0 |
elasticsearch-147171 | elasticsearch | 147171 | false | [CI] OperatorPrivilegesIT testNonOperatorUserCanCallAnalyzeRepositoryAPI failing | Reproduction Line: ./gradlew ':x-pack:plugin:security:qa:operator-privileges-tests:javaRestTest' --tests 'org.elasticsearch.xpack.security.operator.OperatorPrivilegesIT.testNonOperatorUserCanCallAnalyzeRepositoryAPI' -Dtests.seed=7604F88DC268C48D -Dtests.locale=hsb -Dtests.timezone=Europe/London -Druntime.java=26
Appli... | 2a88d3a56c5afdb7a1f30e860a38c0b28ddb3f58 | https://github.com/elastic/elasticsearch/issues/147171 | https://github.com/elastic/elasticsearch/pull/147405 | 2026-04-22T00:00:00 | Major | Concurrency | Hard | true | true | No | true | false | 0 |
etcd-5664 | etcd | 5664 | false | raft: state.commit is out of range | via local-tester with reordering:
```
2016-06-13 17:47:28.006281 I | etcdmain: etcd Version: 3.0.0-beta.0+git
2016-06-13 17:47:28.006332 I | etcdmain: Git SHA: 65e19a1
2016-06-13 17:47:28.006337 I | etcdmain: Go Version: go1.6
2016-06-13 17:47:28.006348 I | etcdmain: Go OS/Arch: linux/amd64
2016-06-13 17:47:28.006353 ... | b07fbbf27cfb2d5d9e209e198eb056c9ff12613f | https://github.com/etcd-io/etcd/issues/5664 | https://github.com/etcd-io/etcd/pull/5690 | 2016-06-14T00:00:00 | Critical | Logic | Hard | true | false | N/A | true | false | 0 |
etcd-14110 | etcd | 14110 | false | panic: unexpected error during txn | ### What happened?
In a Jepsen test run of five etcd 3.5.3 nodes, with process pauses (sending processes SIGSTOP and SIGCONT), compaction (performed via the admin API), and defragmentation (via etcdctl), one etcd process crashed with the following error message:
```
{"level":"panic","ts":"2022-06-13T14:11:27.870Z","c... | v3.5.3 | https://github.com/etcd-io/etcd/issues/14110 | https://github.com/etcd-io/etcd/pull/14149 | 2022-06-13T00:00:00 | Critical | Error handling | Hard | true | false | N/A | true | true | 0 |
etcd-14325 | etcd | 14325 | false | Periodic corruption check doesn’t do anything after compaction | ### What happened?
After turning on `--experimental-corrupt-check-time`, I’ve noticed that any time there has been a compaction and no recent write to etcd, the corruption check fails to actually run.
The output on the leader is
```
{"level":"info","ts":"2022-08-02T17:55:38.557Z","caller":"etcdserver/corrupt.go:... | v3.5.3 | https://github.com/etcd-io/etcd/issues/14325 | https://github.com/etcd-io/etcd/pull/14457 | 2022-08-09T00:00:00 | Major | Logic | Easy | true | false | N/A | true | false | 0 |
etcd-14382 | etcd | 14382 | false | ETCD fails to start after performing alarm list operation and then power off/on | ### What happened?
With a cluster of three nodes setup using https://github.com/bitnami/charts/blob/master/bitnami/etcd/README.md it was noticed that the ETCD nodes failed to start after a forced reboot of the underlying worker nodes. A graceful shutdown will not result in this issue.
The logs indicated a mismatc... | v3.5.4 | https://github.com/etcd-io/etcd/issues/14382 | https://github.com/etcd-io/etcd/pull/14419 | 2022-08-25T00:00:00 | Critical | Logic | Hard | true | false | N/A | true | false | 0 |
etcd-14571 | etcd | 14571 | false | 3.5.5: client auth failures on new member first startup | After cluster upgrade from 3.5.4 to 3.5.5 we are seeing auth issues on new nodes after first startup. Cluster runs in auth-enable mode. New node joins the cluster, fetches the snapshot, but fails all client requests | v3.5.5 | https://github.com/etcd-io/etcd/issues/14571 | https://github.com/etcd-io/etcd/pull/14574 | 2022-10-11T00:00:00 | Critical | Logic | Medium | true | false | null | false | false | 0 |
etcd-14764 | etcd | 14764 | false | Manipulating Keys and Manipulating Users Concurrency Causes Segmentation Violation in Client | ### What happened?
I created a bunch of keys concurrently with goroutines. I created a bunch of users concurrency with goroutines in a separate process.
The process generating the keys aborted with this:
```
{"level":"warn","ts":"2022-11-15T08:33:12.382-0500","logger":"etcd-client","caller":"v3@v3.5.5/retry_int... | v3.5.5 | https://github.com/etcd-io/etcd/issues/14764 | https://github.com/etcd-io/etcd/pull/14770 | 2022-11-15T00:00:00 | Critical | Concurrency | Medium | true | false | N/A | true | true | 0 |
etcd-15247 | etcd | 15247 | false | All leases are revoked when the etcd leader is stuck in handling raft Ready due to slow fdatasync or high CPU | We have 3 etcd nodes which are running on k8s, the etcd pod use portworx storage for db saving, we've found when rebooting the portworx node, it will cause etcd to be stuck in fdatasync, if this etcd node is leader, then it will cause below 2 lease keep alive issues. 1, the old leader still revokes lease after it stepe... | v3.4.3 | https://github.com/etcd-io/etcd/issues/15247 | https://github.com/etcd-io/etcd/pull/16822 | 2023-02-06T00:00:00 | Critical | Logic | Medium | true | true | null | false | false | 0 |
etcd-17507 | etcd | 17507 | true | Etcd not respond to progress notification on newly created watch RPC until there is a event in a watch | If the watch is newly created and not yet synced, and the client sends a progress request during that window, etcd may mark progress as deferred and then never send the progress response unless an actual watch event later occurs. If no events match the watch filter, the client can wait forever for progress notification... | v3.5.10 | https://github.com/etcd-io/etcd/issues/17507 | https://github.com/etcd-io/etcd/pull/17557 | 2024-02-28T00:00:00 | Critical | Logic | Medium | true | true | null | false | false | 0 |
hbase-28376 | hbase | 28376 | true | Master Failed to Start after upgrading from 2.5.10 to 3.0.0-beta-1
2024-12-01T19:00:47,390 ERROR [master/master:16000:becomeActiveMaster] master.HMaster: Failed to become active master
java.lang.IllegalStateException: Expected the service ClusterSchemaServiceImpl [FAILED] to be RUNNING, but the service has FAILED
... | 2306820df8b41d9af5227465ee2cf9e18b8f0b5c | https://issues.apache.org/jira/browse/HBASE-28376 | null | 2024-02-17T00:00:00 | Major | Logic | Hard | false | false | null | false | false | 0 | |
hbase-28830 | hbase | 28830 | false | when a procedure on a table executed as a child procedure, furthur table procedure operations on that table are blocked forever waiting to aquire the table procedure lock .
This issue occur due to not clearing of table lock for the table procedures submitted as child procedures after the changes in HBASE-28683, | rel/3.0.0-beta-1 | https://issues.apache.org/jira/browse/HBASE-28830 | null | 2024-09-10T00:00:00 | Critical | Concurrency | Medium | false | true | Partly Removed | false | false | 0 | |
hbase-28962 | hbase | 28962 | true | Meta replication is inconsistent after startup when reusing HBase storage location for example the new cluster is created with the same S3 storage location that a previous one had.
USE_META_REPLICAS is set to true
META_REPLICAS_NUM is set to 3
For both
hbck output:
ERROR: hbase:meta, replicaId 1 is not found on any r... | db3ba44a4c692d26e70b6030fc519e92fd79f638 | https://issues.apache.org/jira/browse/HBASE-28962 | null | 2024-11-06T00:00:00 | Major | Logic | Hard | true | false | null | true | false | 0 | |
hbase-29021 | hbase | 29021 | false | Symptoms
After upgrading from HBase 2.5.10 to HBase 2.6.1, I observed an inconsistent read across versions.
Before the upgrade, I created a table with one row. Invoking a count command returns 1. After upgrading to 2.6.1, execute the same count command, it returns 0.
count 'test'
Old Version Result:
1 row(s)
Took 0.... | rel/2.6.1 | https://issues.apache.org/jira/browse/HBASE-29021 | null | 2024-12-06T00:00:00 | Critical | Logic | Hard | false | true | Partly Removed | false | false | 0 | |
hbase-29108 | hbase | 29108 | true | Background:
When hbase performs a compaction, it writes the compaction result (1 or more storefiles) to a file in HDFS under /hbase/data/<namespace>/<table>/<region>/.tmp/<columnfamily>/<storefile>
Once the compaction succeeds, the storefile is renamed to /hbase/data/<namespace>/<table>/<region>/<columnfamily>/<storef... | rel/2.5.10 | https://issues.apache.org/jira/browse/HBASE-29108 | null | 2025-02-01T00:00:00 | Major | Error handling | Easy | false | false | N/A | false | false | 0 | |
hdfs-10512 | hdfs | 10512 | true | VolumeScanner may terminate due to NPE in DataNode.reportBadBlocks | VolumeScanner may terminate due to unexpected NullPointerException thrown in DataNode.reportBadBlocks(). This is different from HDFS-8850/HDFS-9190
I observed this bug in a production CDH 5.5.1 cluster and the same bug still persist in upstream trunk.
2016-04-07 20:30:53,830 WARN org.apache.hadoop.hdfs.server.datanode.... | 932aed64d77edcc8483a95c1ce31a4c9ae679446 | https://issues.apache.org/jira/browse/HDFS-10512 | null | 2016-06-09T00:00:00 | Major | Error handling | Medium | false | true | No | false | true | 0 |
hdfs-10733 | hdfs | 10733 | false | NameNode terminated after full GC thinking QJM is unresponsive. | NameNode went into full GC while in AsyncLoggerSet.waitForWriteQuorum(). After completing GC it checks if the timeout for quorum is reached. If the GC was long enough the timeout can expire, and QuorumCall.waitFor() will throw TimeoutExcpetion. Finally FSEditLog.logSync() catches the exception and terminates NameNode. | 9130af3f74a7923fff3f80360ba043ab2c6558a9 | https://issues.apache.org/jira/browse/HDFS-10733 | null | 2016-08-08T00:00:00 | Major | Error handling | Medium | false | false | N/A | false | false | 0 |
hdfs-11583 | hdfs | 11583 | false | Parent spans are not initialized to NullScope for every DFSPacket | Each packet received by the run() method of DataStreamer class, uses the parents field of the DFSPacket to create a new dataStreamer span, which in turn creates a writeTo span as its child span. The parents field is initialized when the packet is added to the dataQueue and the value is initialized from the ThreadLocal.... | release-2.7.1 | https://issues.apache.org/jira/browse/HDFS-11583 | null | 2017-03-27T00:00:00 | Major | Logic | Easy | false | true | No | false | false | 0 |
hdfs-11701 | hdfs | 11701 | false | NPE from Unresolved Host causes permanent DFSInputStream failures | We recently encountered the following NPE due to the DFSInputStream storing old cached block locations from hosts which could no longer resolve.
Caused by: java.lang.NullPointerException
at org.apache.hadoop.hdfs.DFSClient.isLocalAddress(DFSClient.java:1122)
at org.apache.hadoop.hdfs.shortcircuit.DomainSocketFactory.ge... | 456705a07c8b80658950acc99f23086244c6b20f | https://issues.apache.org/jira/browse/HDFS-11701 | null | 2017-04-25T00:00:00 | Major | Error handling | Medium | false | false | N/A | false | true | 0 |
hdfs-13051 | hdfs | 13051 | true | Fix dead lock during async editlog rolling if edit queue is full | when doing rolleditlog it acquires fs write lock，then acquire FSEditLogAsync lock object，and write 3 EDIT(the second one override logEdit method and return true)
in extremely case，when FSEditLogAsync's logSync is very slow，editPendingQ(default size 4096)is full，it case IPC thread can not offer edit object into editPen... | 96892c469b16c5aaff1b7c42f66f820344256bc2 | https://issues.apache.org/jira/browse/HDFS-13051 | https://issues.apache.org/jira/secure/attachment/12909545/deadlock.patch | 2018-01-23T00:00:00 | Major | Concurrency | Hard | true | false | N/A | false | true | 0 |
hdfs-13101 | hdfs | 13101 | false | Yet another fsimage corruption related to snapshot | FSImage corruption happened after HDFS snapshots were taken. Cluster was not used
at that time.
When namenode restarts it reported NULL pointer exception:
15/11/07 10:01:15 INFO namenode.FileJournalManager: Recovering unfinalized segments in /tmp/fsimage_checker_5857/fsimage/current
15/11/07 10:01:15 INFO namenode.FSIm... | 167acd87daf61be9a27dfb5af6e7cbacb96e44de | https://issues.apache.org/jira/browse/HDFS-13101 | https://issues.apache.org/jira/secure/attachment/12977155/HDFS-13101.004.patch | 2018-02-01T00:00:00 | Critical | Logic | Medium | true | false | N/A | true | false | 0 |
hdfs-13179 | hdfs | 13179 | false | TestLazyPersistReplicaRecovery#testDnRestartWithSavedReplicas fails intermittently | The error caused by TimeoutException because the test is waiting to ensure that the file is replicated to DISK storage but the replication can't be finished to DISK during the 30s timeout in ensureFileReplicasOnStorageType(), but the file is still on RAM_DISK - so there is no data loss.
Adding the following to TestLazy... | 5abd0148eb9484381c7ff92b4bfab9ac8a81c681 | https://issues.apache.org/jira/browse/HDFS-13179 | https://issues.apache.org/jira/secure/attachment/12991953/HDFS-13179.003.patch | 2018-02-22T00:00:00 | Critical | Logic | Easy | true | true | No | false | true | 0 |
hdfs-13292 | hdfs | 13292 | false | Crypto command should give proper exception when trying to set key on existing EZ directory | Scenario:
Create a Dir
Create EZ for the above dir with Key1
Again you can try to create ZONE for same directory with Diff Key i.e Key2
hadoopclient> hadoop key list
Listing keys for KeyProvider: org.apache.hadoop.crypto.key.kms.LoadBalancingKMSClientProvider@152aa092
key2
key1
hadoopclient> hdfs dfs -mkdir /kms
hadoop... | 00ebec89f101347a5da44657e388b30c57ed9deb | https://issues.apache.org/jira/browse/HDFS-13292 | null | 2018-03-15T00:00:00 | Major | Error handling | Medium | false | false | N/A | false | false | 0 |
hdfs-13893 | hdfs | 13893 | false | DiskBalancer: no validations for Disk balancer commands | 1 Run the Disk Balancer commands with extra arguments passing
hadoopclient> hdfs diskbalancer -plan hostname --thresholdPercentage 2 *sgfsdgfs*
2018-08-31 14:57:35,454 INFO planner.GreedyPlanner: Starting plan for Node : hostname:50077
2018-08-31 14:57:35,457 INFO planner.GreedyPlanner: Disk Volume set fb67f00c-e33... | 98d20656433cdec76c2108d24ff3b935657c1e80 | https://issues.apache.org/jira/browse/HDFS-13893 | https://issues.apache.org/jira/secure/attachment/12949642/HDFS-13893.003.patch | 2018-09-04T00:00:00 | Major | Error handling | Easy | false | false | N/A | false | false | 0 |
hdfs-15175 | hdfs | 15175 | true | Multiple CloseOp shared block instance causes the standby namenode to crash when rolling editlog | Crash exception
2020-02-16 09:24:46,426 [507844305] - ERROR [Edit log tailer:FSEditLogLoader@245] - Encountered exception on operation CloseOp [length=0, inodeId=0, path=..., replication=3, mtime=1581816138774, atime=1581814760398, blockSize=536870912, blocks=[blk_5568434562_4495417845], permissions=da_music:hdfs:rw-r-... | 3c8a48e681babe4038e49205624e7be183b3466e | https://issues.apache.org/jira/browse/HDFS-15175 | https://issues.apache.org/jira/secure/attachment/13005703/HDFS-15175-trunk.1.patch | 2018-02-22T00:00:00 | Critical | Logic | Medium | true | false | N/A | true | false | 0 |
ignite-4205 | ignite | 4205 | false | CassandraCacheStore.loadCache() method starts a generic thread pool for parallel data load. Threads in this thread pool can't deserialize Ignite internal objects (e.g. IgniteKernal) which can cause unexpected behavior. Here is one of the scenarios:
There is column in Cassandra which stores an object as BLOB using ... | 90717b4b7230ca5f5e0fd9318d0aa0a6f5e48064 | https://issues.apache.org/jira/browse/IGNITE-4205 | null | 2016-11-10T00:00:00 | Major | Logic | Medium | false | false | null | false | false | 0 | |
ignite-6860 | ignite | 6860 | true | Having the stack trace
Caused by: org.apache.ignite.binary.BinaryObjectException: Failed to marshal object with optimized marshaller: [org.apache.logging.log4j.core.config.AppenderControl@302e61a8]
at org.apache.ignite.internal.binary.BinaryWriterExImpl.marshal0(BinaryWriterExImpl.java:186)
at org.apac... | 7366809ed0ea7a943e1c80079fdbeb556ee86245 | https://issues.apache.org/jira/browse/IGNITE-6860 | null | 2017-11-10T00:00:00 | Major | Error handling | Hard | false | false | null | false | true | 0 | |
ignite-9133 | ignite | 9133 | true | In the cases when DHT transaction is empty (i.e. no keys were enlisted) after the enlist step, we need to rollback local DHT transaction to exclude it from the further transaction flow in order to performance increasing.
An ordinary Dht tx rollback GridDhtTxLocal#rollbackDhtLocalAsync is not suitable in this situation... | 824939ba1f238621bd89a5ff445c6932c79e5db9 | https://issues.apache.org/jira/browse/IGNITE-9133 | null | 2018-07-30T00:00:00 | Major | Logic | Medium | true | false | null | true | false | 0 | |
ignite3-15596 | ignite3 | 15596 | false | Calcite. java.lang.Integer cannot be cast to java.lang.Long with JOIN and EXISTS.
Export
statement ok
CREATE TABLE tbl(a TINYINT, b SMALLINT, c INTEGER, d BIGINT, e VARCHAR, f DATE, g TIMESTAMP)
statement ok
INSERT INTO tbl VALUES (1, 2, 3, 4, '5', DATE '1992-01-01', TIMESTAMP '1992-01-01 00:00:00')
query I
SELE... | ed519fb2fa2a89e11a8ab7fad3069f7e1f7d74b3 | https://issues.apache.org/jira/browse/IGNITE-15596 | null | 2021-09-24T00:00:00 | Major | Logic | Easy | false | false | null | false | false | 0 | |
ignite3-15724 | ignite3 | 15724 | false | MetastorageManager uses incorrect predicate in order to create corresponding raft group service Currently, when the MetaStorageManager is started, the metastorage raft group is reduced to online nodes that satisfies to the metaStorageNodesContainsLocPred.
String[] metastorageNodes = this.locCfgMgr.configurationRe... | 2abd5b9f5525ac2ea9314e45cea9deb6be871418 | https://issues.apache.org/jira/browse/IGNITE-15724 | null | 2021-10-11T00:00:00 | Critical | Logic | Medium | false | false | null | false | false | 0 | |
ignite3-17428 | ignite3 | 17428 | false | Current version of TableManager#tableAsyncInternal can possibly not detect table that is being created while tableAsyncInternal is called. Scenario:
tableAsyncInternal checks tablesByIdVv.latest() and there is no table
the table creation started, table metadata appears in meta storage
TableEvent.CREATE is ... | 2c4d07597235ef54b23c9726a755e5876b98e1a1 | https://issues.apache.org/jira/browse/IGNITE-17428 | null | 2022-07-27T00:00:00 | Major | Logic | Medium | false | true | No | false | false | 0 | |
ignite3-20498 | ignite3 | 20498 | false |
Prevent potential catalog version order violations.
Currently, catalog versions are stored in an ordered structure. The activation timestamp is used as a key, which depends on the configuration property "schemaSync.delayDuration".
Changes to "delayDuration" parameter in runtime may lead to a violation of the order in... | 7a1540365d89d188da1f0d22c27c96ddc21485ac | https://issues.apache.org/jira/browse/IGNITE-20498 | null | 2023-09-26T00:00:00 | Major | Logic | Medium | false | false | null | false | false | 0 | |
ignite3-21776 | ignite3 | 21776 | false | "Create table if not exists" not a thread safe : In two parallel clients:
1) Run DDL (using two clients):
```
create table if not exists parallelCreateTable0
(id INTEGER not null,
val VARCHAR not null,
primary key (id))
```
2) Get parallelCreateTable0 using ignite.tables() API or select from jdbc
Expected resu... | b08fbb60872f5c88209152992bce392f92109f87 | https://issues.apache.org/jira/browse/IGNITE-21776 | null | 2024-03-18T00:00:00 | Major | Logic | Medium | false | false | null | false | false | 0 | |
ignite3-21827 | ignite3 | 21827 | true | Union returns incorrect result for NUMERIC/DECIMAL and INT columns
Simple reproducer:
sql("CREATE TABLE t0(id INT PRIMARY KEY, val INT)");
sql("CREATE TABLE t1(id INT PRIMARY KEY, val DECIMAL)");
sql("INSERT INTO t0 VALUES (1, 10)");
sql("INSERT INTO t1 VALUES (1, 10)");
assertQuery("SELECT id, v... | 166ade294eb8311cb0dc46bcdcedb3620bea7eb9 | https://issues.apache.org/jira/browse/IGNITE-21827 | null | 2024-03-22T00:00:00 | Major | Logic | Medium | false | false | N/A | false | false | 0 | |
ignite3-22597 | ignite3 | 22597 | true | Motivation
The trace id is generated at the place where the original exception was created and should be propogated to other derevatives. Right now, we send only an exception message and class name to the RAFT listener response.
private void sendSMError(RpcContext ctx, Throwable th, boolean compacted) {
RpcReques... | 787e868ede5a489b49c4cd9d58d3c467b10e46ed | https://issues.apache.org/jira/browse/IGNITE-22597 | null | 2024-06-27T00:00:00 | Major | Error handling | Easy | false | false | null | false | false | 0 | |
kafka-6446 | kafka | 6446 | true | KafkaProducer with transactionId endless waits when bootstrap server is down | When bootstrap server is down, a KafkaProducer with transactionId endless waits on initTransactions.
The timeouts don't apply to that operation: don't honor the TRANSACTION_TIMEOUT_CONFIG.
Attached an example of my code to reproduce the scenario.
import java.util.HashMap;
import java.util.Map;
import org.apache.ka... | 1.0.0 | https://issues.apache.org/jira/browse/KAFKA-6446 | https://github.com/apache/kafka/pull/4563 | 2018-01-15T00:00:00 | Critical | Concurrency | Medium | true | false | N/A | false | false | 0 |
End of preview. Expand in Data Studio
DistDebug-Bench
DistDebug-Bench evaluates whether LLM agents can find the root cause of real-world bugs in distributed systems. It contains 120 resolved bugs from 11 open-source systems. For each bug, an agent receives the original issue report, any attached logs and screenshots, and a checkout of the system's source at the pre-fix snapshot. It must name the responsible files, functions, and line ranges. Diagnoses are scored against expert-annotated root-cause blocks.
- Paper: DistDebug-Bench: Can LLM Agents Diagnose Bugs in Distributed Systems? (NeurIPS 2026, Track on Evaluations and Datasets)
- Code, harness, and evaluator: https://github.com/OrderLab/distdebug-bench
- Website and leaderboard: https://distdebug.com
- Agent traces behind the paper's Table 1: https://huggingface.co/datasets/OrderLab/distdebug-bench-traces
- Version: 1.0.0
Contents
| Config / path | Rows | Description |
|---|---|---|
instances (data/instances.jsonl) |
120 | One row per bug: issue report, pre-fix snapshot, links, and metadata. |
gold_blocks (data/gold_blocks.jsonl) |
218 | One row per gold root-cause block, keyed by instance_id. |
systems (data/systems.jsonl) |
11 | Source systems with repository URLs and upstream licenses. |
artifacts/<instance_id>/ |
17 screenshots, 4 logs | Files attached to the issue reports of 7 bugs. |
croissant.json describes all of the above, including the attachments, in
Croissant format with Responsible AI
fields.
What the agent sees
The task input is title, description, the bug's attachments, and the
system's repository checked out at branch. The remaining columns
(issue_url, fix_url, and the characterization labels) are
for analysis and must not be shown to the agent: the links lead to the fix.
The harness in the GitHub repository prepares workspaces with anonymized paths
and without commits after the snapshot.
Gold blocks
Each block gives a repository-relative file, the enclosing function, and a
start_line/end_line range in the pre-fix snapshot, scored as half-open
[start_line, end_line). Blocks capture the annotated causal implementation
established through issue resolution, not every location where the bug could
be repaired. Score diagnoses with the evaluator in the GitHub repository; it
reproduces the paper's scores and reports file-, function-, line-, and
block-level coverage and precision, plus strict block coverage and mention
coverage.
DistDebug-Bench Lite
DistDebug-Bench Lite is a 34-bug subset (lite == true), selected for
report and label quality and for lower evaluation cost. Report Lite results
separately from full-benchmark results.
Loading
from datasets import load_dataset
instances = load_dataset("OrderLab/distdebug-bench", "instances", split="test")
gold = load_dataset("OrderLab/distdebug-bench", "gold_blocks", split="test")
lite = instances.filter(lambda row: row["lite"])
With mlcroissant, every record set, including the screenshots and logs, can be loaded from the metadata alone:
import mlcroissant as mlc
ds = mlc.Dataset("https://huggingface.co/datasets/OrderLab/distdebug-bench/resolve/main/croissant.json")
for record in ds.records("artifact_logs"):
print(record["artifact_logs/instance_id"], record["artifact_logs/file_name"])
Systems and licenses
| System | Language | Issue tracker | Upstream license | Bugs | Lite |
|---|---|---|---|---|---|
| Elasticsearch | java | GitHub Issues | AGPL-3.0-only OR SSPL-1.0 OR Elastic-2.0 |
19 | 4 |
| etcd | go | GitHub Issues | Apache-2.0 |
8 | 1 |
| HBase | java | Apache JIRA | Apache-2.0 |
5 | 3 |
| HDFS | java | Apache JIRA | Apache-2.0 |
10 | 3 |
| Ignite | java | Apache JIRA | Apache-2.0 |
3 | 2 |
| Ignite 3 | java | Apache JIRA | Apache-2.0 |
7 | 2 |
| Kafka | java | Apache JIRA | Apache-2.0 |
8 | 8 |
| Kubernetes | go | GitHub Issues | Apache-2.0 |
10 | 1 |
| Ray | python | GitHub Issues | Apache-2.0 |
21 | 4 |
| TiKV | rust | GitHub Issues | Apache-2.0 |
11 | 3 |
| ZooKeeper | java | Apache JIRA | Apache-2.0 |
18 | 3 |
The annotations, metadata, and code in this release are under the MIT license. Issue titles, descriptions, and attachments are reproduced from public issue trackers and remain subject to the terms of those trackers and projects. The dataset references paths and line numbers in the upstream repositories but does not redistribute their source code; the licenses above are those of the repositories' current default branches, and historical snapshots may differ.
Limitations
The benchmark covers resolved bugs in public, open-source distributed systems
and evaluates code-level diagnosis only. Selection favors bugs whose fixes
support precise labeling. Because the issues are public, models may have seen
them during pretraining; the paper reports leakage controls and contamination
probes. See the paper and croissant.json for the full discussion.
Citation
@inproceedings{jiang2026distdebugbench,
title = {{DistDebug-Bench}: Can {LLM} Agents Diagnose Bugs in Distributed Systems?},
author = {Jiang, Yuxuan and Liu, Chenglin and Rose, John and Miller, Mason and Thavamani, Ruthesh and Wu, Juntao and Huang, Yinwei and Xu, Jiahao and Huang, Peng},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS), Track on Evaluations and Datasets},
year = {2026}
}
- Downloads last month
- 18