[
https://issues.apache.org/jira/browse/KUDU-3807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18117859#comment-18117859
]
Abhishek Chennaka commented on KUDU-3807:
-----------------------------------------
The sys-catalog copy fails and the Kudu master dies:
{code:java}
I20260921 18:47:56.555657 10274 master_runner.cc:316] Copying system tablet
from 127.20.2.125:41893
I20260921 18:47:56.740595 10274 tablet_copy_client.cc:323] T
00000000000000000000000000000000 P 5b20771fa90c4e2b9a52a2f314505255: tablet
copy: Beginning tablet copy session from remote peer at address
127.20.2.125:41893
Remote error: unable to begin tablet copy session: Error beginning tablet copy
session for tablet 00000000000000000000000000000000: UNKNOWN_ERROR: received
error code Not found: Unable to open block file for block 17394072812956168263:
Can't find block when opening log block: 17394072812956168263 from remote
service{code}
This is due to a minor compaction run which deleted the needed block:
{code:java}
I20260921 18:47:56.914476 4620 maintenance_manager.cc:643] P
e83b2dcf51f8438d85d09c78feef0625:
MinorDeltaCompactionOp(00000000000000000000000000000000) complete. Timing: real
0.052s user 0.009s sys 0.000s Metrics:
{"cfile_cache_miss":4,"cfile_cache_miss_bytes":1194,"cfile_init":1,"delete_count":0,"delta_blocks_compacted":2,"delta_iterators_relevant":2,"dirs.queue_time_us":138,"dirs.run_cpu_time_us":141,"dirs.run_wall_time_us":674,"lbm_read_time_us":251,"lbm_reads_lt_1ms":8,"lbm_write_time_us":238,"lbm_writes_lt_1ms":4,"reinsert_count":0,"update_count":28}{code}
We probably have to either ensure that the tablet-copy source session pins the
blocks it snapshots (or re-read the superblock / tolerate a block that a
concurrent compaction has retired) instead of returning NotFound or retry the
sys-catalog copy in such scenarios.
> Kudu3762Test.AddMaster fails from time to time
> ----------------------------------------------
>
> Key: KUDU-3807
> URL: https://issues.apache.org/jira/browse/KUDU-3807
> Project: Kudu
> Issue Type: Bug
> Components: test
> Affects Versions: 1.19.0
> Reporter: Alexey Serbin
> Priority: Major
> Attachments: external_mini_cluster-test.asan.20260921.txt.xz
>
>
> The test scenario added in the scope of KUDU-3762 fails from time to time.
> At least, it failed in ASAN build config (no sanitizer warnings) with an
> error like below:
> {noformat}
> src/kudu/mini-cluster/external_mini_cluster-test.cc:490: Failure
> Failed
> Bad status: Timed out: Timed out after 60s waiting for master
> (127.20.2.124:39633) startup
> {noformat}
> Full log is attached.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
