Repository navigation
Antalya 26.8 Backport of #131963 - Clear --jwt and the other credential options from the command line of clickhouse-client and clickhouse-keeper-client - #2885
Conversation
…ptions-from-command-line Clear `--jwt` and the other credential options from the command line of `clickhouse-client` and `clickhouse-keeper-client`
… --oauth-client-secret clickhouse-client in antalya-26.8 has no --oauth-client-secret option, so the client rejected it and exited before connecting, failing the test.
|
CI fix (fixed), asked for in Port-AL by zvonand: Adapted 05813_clear_credentials_from_command_line to 26.8: dropped --oauth-client-secret, which the 26.8 client lacks; test passes locally in 1.4s Pushed 42af140 on top of 0e5d90c, for the failures blau-ai's analysis of 0e5d90c blamed on this PR: |
CI triage for
|
| Check | Classification | PR-caused? |
|---|---|---|
| Build (wasm_parser) | Infra: no runner was ever assigned | No |
| SQLLogic test | Infra: job hit its 3 h timeout | No |
Integration tests (amd_asan_ubsan, db disk, old analyzer, 3/8): test_s3_cluster/test.py::test_graceful_shutdown |
Flaky, timing-sensitive swarm shutdown test | No (high confidence) |
| RegressionTestsRelease / Swarms | Known issue #2896 | No |
| Grype Scan keeper / server-alpine | Image CVEs, not code | No |
| DCO | Sign-off policy on the backported upstream commits | No (process) |
1. Build (wasm_parser): infra
The GitHub job 134782092682 reports conclusion=cancelled, runner_name="", steps=0, and ran 2026-11-18T11:52 → 2026-11-19T11:52 (exactly 24 h). It was queued for the altinity-func-tester-aarch64 on-demand pool, never got a runner, and was cancelled at the 24 h limit. Nothing was compiled. The same job is also stuck pending on #2912.
Next step: re-run when that runner pool is available. No code change needed.
2. SQLLogic test: timeout
[2026-11-19 01:00:00] WARNING: Timeout exceeded [10800] for [4977]
[2026-11-19 01:00:13] Error response from daemon: cannot remove container "praktika_d7faa2714e6c": could not kill container ...
[2026-11-19 01:00:30] WARNING: Job timed out: [SQLLogic test], timeout [10800], exit code [-15]
The job was killed during complete-test, so no test-level result was produced. For comparison, the last full SQLLogic run on antalya-26.8 (MasterCI on cafe527, 2026-11-14) passed but took 8475 s (2 h 21 m) of the 10800 s budget. That leaves little headroom, and a slow runner (the container could not even be killed afterwards) is enough to push it over. A change to client-side argv scrubbing has no plausible effect on the server's SQL execution speed.
Next step: re-run.
3. Integration tests (amd_asan_ubsan, db disk, old analyzer, 3/8): test_s3_cluster/test.py::test_graceful_shutdown, flaky
test_s3_cluster/test.py:2171: in test_graceful_shutdown
assert errors == 0
E assert 1 == 0 (retry: assert 2 == 0)
1 failure out of 732 tests; the other 731 passed. It also failed its retry, so I'm basing the verdict on what the test does, not just the low count:
- The test runs 10 concurrent
s3Clusterqueries, issuesSYSTEM STOP SWARM MODEon one replica, waits a fixedtime.sleep(15), then kill -9's that replica. It counts any query that errored or returned a partial result. - The 15 s drain is a hand-tuned timeout. It was already lengthened once for sanitizer slowness (
287493fe9c4 "...lengthen swarm shutdown drain", comment: "TSan is several times slower, so 3s is not enough"). This run is ASan+UBSan, where an in-flight object can still be running when the kill lands. - Nothing in the PR touches swarm,
s3Clusteror server shutdown code. The only client-side change zeroes argv after option parsing. If that broke query execution, everynode.query(...)in the suite would fail, but 731 other tests in this same job ran queries through the same client and passed. - I didn't find the exception text in the job log, because the test swallows
QueryRuntimeExceptioninto theerrorscounter. The server logs are inlogs.tar.gzof that job if someone wants to confirm which query failed.
Next step: re-run. If it keeps recurring on 26.8, the fix belongs in the test (poll system.processes on the stopped node until it's drained instead of using a fixed 15 s sleep), not in this PR.
4. RegressionTestsRelease / Swarms: known issue #2896
2 of 1535 scenarios failed: /swarms/feature/node failure/cpu overload and .../cpu overload all swarm nodes:
Code: 209. DB::Exception: ... Timeout exceeded while reading from socket (socket (172.18.0.10:9000), receive timeout 300000 ms): While executing Remote. (SOCKET_TIMEOUT)
This is the known issue tracked in #2896, "Antalya 26.8: Swarm query hangs for receive_timeout (300 s) when swarm nodes are frozen". It's a pre-existing base-branch issue and unrelated to client argv handling.
5. Grype Scan (clickhouse-keeper, clickhouse-server-alpine): image CVEs
Both reports list the same image-level findings (CVE-2026-95691, CVE-2026-95692, CVE-2026-88822, CVE-2026-88823, CVE-2026-107496, plus a medium CVE-2025-60876), all matched via nvd:cpe. The regular (non-alpine) server image scan passes on the same build. This PR changes no Dockerfiles or dependencies, so these come from the base images or packages. Handle them in the docker or base-image track, not here.
6. DCO: process
The backport carries upstream's merge commit 0e5d90c318, which has no Signed-off-by from an Altinity author. The same check is red on the sibling backports #2901 and #2910. This is the usual state for backport PRs here, so it's a maintainer decision, not something to fix in this PR's code.
Bottom line: nothing to fix in the PR itself. I'd re-run SQLLogic, Build (wasm_parser) and Integration tests (amd_asan_ubsan, db disk, old analyzer, 3/8). Treat Swarms (#2896), Grype and DCO as pre-existing or non-blocking.
Changelog category (leave one):
Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
clickhouse-clientnow clears the values of--jwt,--ssh-key-passphrase,--one-time-passwordand--oauth-client-secretfrom its command line, as it already did for--password, so they are no longer visible inpsand/proc/<pid>/cmdline.clickhouse-keeper-clientnow does the same for--passwordand--identity. Closes ClickHouse#131191. (ClickHouse#131963 by @groeneai)Documentation entry for user-facing changes
...
CI/CD Options
Exclude tests:
Regression jobs to run: