[jira] [Commented] (KUDU-3549) String gauge exposed in prometheus format
[
https://issues.apache.org/jira/browse/KUDU-3549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815376#comment-17815376
]
ASF subversion and git services commented on KUDU-3549:
---
Commit 81823c1ed73d2d57ab08ad30098a96ab66f70b49 in kudu's branch
refs/heads/branch-1.17.x from Alexey Serbin
[ https://gitbox.apache.org/repos/asf?p=kudu.git;h=81823c1ed ]
KUDU-3549 fix WriteAsPrometheus() for non-arithmetic gauges
This patch fixes WriteAsPrometheus() implementation for string-based
gauges. The original changelist that introduced Prometheus metrics
had a proper implementation of WriteAsPrometheus() only for StringGauge,
but any FunctionGauge for a non-arithmetic type (e.g., for std::string)
would still output a string value in Prometheus text format, and that
could not be consumed by Prometheus. The patch also contains a test
scenario that would fail without the fix.
This is a follow-up to 00efc6826ac9a1f5d10750296c7357790a041fec.
Change-Id: Ib7128f52729c7f984004811153a7eecc8ffe751b
Reviewed-on: http://gerrit.cloudera.org:8080/20990
Tested-by: Alexey Serbin
Reviewed-by: Marton Greber
Reviewed-by: Attila Bukor
(cherry picked from commit 6b1c1eb0c97a2349e0b3fa098bf40f8147b43a60)
Conflicts:
src/kudu/util/metrics.cc
src/kudu/util/metrics.h
Reviewed-on: http://gerrit.cloudera.org:8080/21007
Reviewed-by: Yingchun Lai
> String gauge exposed in prometheus format
> -
>
> Key: KUDU-3549
> URL: https://issues.apache.org/jira/browse/KUDU-3549
> Project: Kudu
> Issue Type: Bug
> Components: metrics
>Affects Versions: 1.17.0
>Reporter: YUBI LEE
>Assignee: Alexey Serbin
>Priority: Major
>
> According to KUDU-3375, "Kudu now exposes all its metrics except for string
> gauges in Prometheus format via the embedded webserver's
> `/metrics_prometheus` endpoint".
>
> *
> [https://github.com/apache/kudu/blob/89e2715faf96afe0b67482166fda9c8699e8052f/docs/prior_release_notes.adoc?plain=1#L143-L145]
> * https://issues.apache.org/jira/browse/KUDU-3375
>
> However, with this commit
> ([https://github.com/apache/kudu/commit/e65ea38a4860c007d93ada9c991bccec903a80b1)]
> , string gauge related to clock_ntp_status is exposed.
>
> {code:java}
> # HELP kudu_master_clock_ntp_status Output of ntp_adjtime()/ntp_gettime()
> kernel API call
> # TYPE kudu_master_clock_ntp_status gauge
> kudu_master_clock_ntp_status{unit_type="state"} now:1706665936956760
> maxerror:70013 status:ok {code}
> It prevents prometheus operators from collecting prometheus metrics for kudu.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Commented] (KUDU-3433) ClientTest.TestDeleteWithDeletedTableReserveSecondsWorks is flaky
[
https://issues.apache.org/jira/browse/KUDU-3433?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815380#comment-17815380
]
Alexey Serbin commented on KUDU-3433:
-
The test failing from time to time. The logs from one of the recent failures
attached. [^client-test.3.txt.xz]
> ClientTest.TestDeleteWithDeletedTableReserveSecondsWorks is flaky
> -
>
> Key: KUDU-3433
> URL: https://issues.apache.org/jira/browse/KUDU-3433
> Project: Kudu
> Issue Type: Bug
>Affects Versions: 1.17.0
>Reporter: Alexey Serbin
>Priority: Major
> Attachments: client-test.2.txt.xz, client-test.3.txt.xz,
> client-test.4.txt.xz
>
>
> The {{TestDeleteWithDeletedTableReserveSecondsWorks}} in {{client-test}}
> sometimes fails with the following message:
> {noformat}
> src/kudu/client/client-test.cc:5436: Failure
> Value of: tables.empty()
>
> Actual: false
>
> Expected: true
> {noformat}
> I'm attaching a full log for reference.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Updated] (KUDU-3433) ClientTest.TestDeleteWithDeletedTableReserveSecondsWorks is flaky
[
https://issues.apache.org/jira/browse/KUDU-3433?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Serbin updated KUDU-3433:
Attachment: client-test.3.txt.xz
> ClientTest.TestDeleteWithDeletedTableReserveSecondsWorks is flaky
> -
>
> Key: KUDU-3433
> URL: https://issues.apache.org/jira/browse/KUDU-3433
> Project: Kudu
> Issue Type: Bug
>Affects Versions: 1.17.0
>Reporter: Alexey Serbin
>Priority: Major
> Attachments: client-test.2.txt.xz, client-test.3.txt.xz,
> client-test.4.txt.xz
>
>
> The {{TestDeleteWithDeletedTableReserveSecondsWorks}} in {{client-test}}
> sometimes fails with the following message:
> {noformat}
> src/kudu/client/client-test.cc:5436: Failure
> Value of: tables.empty()
>
> Actual: false
>
> Expected: true
> {noformat}
> I'm attaching a full log for reference.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Resolved] (KUDU-3549) String gauge exposed in prometheus format
[
https://issues.apache.org/jira/browse/KUDU-3549?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Serbin resolved KUDU-3549.
-
Fix Version/s: 1.18.0
1.17.1
Resolution: Fixed
> String gauge exposed in prometheus format
> -
>
> Key: KUDU-3549
> URL: https://issues.apache.org/jira/browse/KUDU-3549
> Project: Kudu
> Issue Type: Bug
> Components: metrics
>Affects Versions: 1.17.0
>Reporter: YUBI LEE
>Assignee: Alexey Serbin
>Priority: Major
> Fix For: 1.18.0, 1.17.1
>
>
> According to KUDU-3375, "Kudu now exposes all its metrics except for string
> gauges in Prometheus format via the embedded webserver's
> `/metrics_prometheus` endpoint".
>
> *
> [https://github.com/apache/kudu/blob/89e2715faf96afe0b67482166fda9c8699e8052f/docs/prior_release_notes.adoc?plain=1#L143-L145]
> * https://issues.apache.org/jira/browse/KUDU-3375
>
> However, with this commit
> ([https://github.com/apache/kudu/commit/e65ea38a4860c007d93ada9c991bccec903a80b1)]
> , string gauge related to clock_ntp_status is exposed.
>
> {code:java}
> # HELP kudu_master_clock_ntp_status Output of ntp_adjtime()/ntp_gettime()
> kernel API call
> # TYPE kudu_master_clock_ntp_status gauge
> kudu_master_clock_ntp_status{unit_type="state"} now:1706665936956760
> maxerror:70013 status:ok {code}
> It prevents prometheus operators from collecting prometheus metrics for kudu.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Commented] (KUDU-3549) String gauge exposed in prometheus format
[
https://issues.apache.org/jira/browse/KUDU-3549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815382#comment-17815382
]
Alexey Serbin commented on KUDU-3549:
-
[~eub],
Thank you very much for reporting the issue!
The bug has been fixed.
The fix are available in both the 1.17.x and the main branches of the Kudu's
git repo.
> String gauge exposed in prometheus format
> -
>
> Key: KUDU-3549
> URL: https://issues.apache.org/jira/browse/KUDU-3549
> Project: Kudu
> Issue Type: Bug
> Components: metrics
>Affects Versions: 1.17.0
>Reporter: YUBI LEE
>Assignee: Alexey Serbin
>Priority: Major
> Fix For: 1.18.0, 1.17.1
>
>
> According to KUDU-3375, "Kudu now exposes all its metrics except for string
> gauges in Prometheus format via the embedded webserver's
> `/metrics_prometheus` endpoint".
>
> *
> [https://github.com/apache/kudu/blob/89e2715faf96afe0b67482166fda9c8699e8052f/docs/prior_release_notes.adoc?plain=1#L143-L145]
> * https://issues.apache.org/jira/browse/KUDU-3375
>
> However, with this commit
> ([https://github.com/apache/kudu/commit/e65ea38a4860c007d93ada9c991bccec903a80b1)]
> , string gauge related to clock_ntp_status is exposed.
>
> {code:java}
> # HELP kudu_master_clock_ntp_status Output of ntp_adjtime()/ntp_gettime()
> kernel API call
> # TYPE kudu_master_clock_ntp_status gauge
> kudu_master_clock_ntp_status{unit_type="state"} now:1706665936956760
> maxerror:70013 status:ok {code}
> It prevents prometheus operators from collecting prometheus metrics for kudu.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Commented] (KUDU-3535) [tserver] Should clear log cache while tombstoning a replica.
[ https://issues.apache.org/jira/browse/KUDU-3535?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815418#comment-17815418 ] ASF subversion and git services commented on KUDU-3535: --- Commit 368225e87f77851f8cdf98fc4b7670aaac7a773e in kudu's branch refs/heads/master from 宋家成 [ https://gitbox.apache.org/repos/asf?p=kudu.git;h=368225e87 ] KUDU-3535 Clear log cache while tombstoning a tablet replica. The log cache of a replica still exists even if the replica has been already tombstoned. This problem might take place if we decrease the replication factor of a table with high throughput. So we should clear the log cache while deleting the replica with delete type TABLET_DATA_TOMBSTONED. Change-Id: I6cf545e604f80d41e7ebd9660acfd2e928cd27a9 Reviewed-on: http://gerrit.cloudera.org:8080/20822 Reviewed-by: Alexey Serbin Tested-by: Alexey Serbin > [tserver] Should clear log cache while tombstoning a replica. > - > > Key: KUDU-3535 > URL: https://issues.apache.org/jira/browse/KUDU-3535 > Project: Kudu > Issue Type: Bug > Components: tserver >Reporter: Song Jiacheng >Priority: Major > Attachments: image-2023-12-20-15-04-14-249.png, > image-2023-12-20-15-05-02-970.png, image-2023-12-20-15-05-21-394.png > > > The log cache of a replica still exists even if the replica has been > tombstoned. The 2 pictures below show the problem. > !image-2023-12-20-15-05-02-970.png|width=372,height=171! > !image-2023-12-20-15-05-21-394.png|width=369,height=184! > We should clear the log cache while delete the replica with delete type " > TABLET_DATA_TOMBSTONED" -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KUDU-3549) String gauge exposed in prometheus format
[
https://issues.apache.org/jira/browse/KUDU-3549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815442#comment-17815442
]
YUBI LEE commented on KUDU-3549:
[~aserbin] ,
Thank you for resolving this issue very quickly!
> String gauge exposed in prometheus format
> -
>
> Key: KUDU-3549
> URL: https://issues.apache.org/jira/browse/KUDU-3549
> Project: Kudu
> Issue Type: Bug
> Components: metrics
>Affects Versions: 1.17.0
>Reporter: YUBI LEE
>Assignee: Alexey Serbin
>Priority: Major
> Fix For: 1.18.0, 1.17.1
>
>
> According to KUDU-3375, "Kudu now exposes all its metrics except for string
> gauges in Prometheus format via the embedded webserver's
> `/metrics_prometheus` endpoint".
>
> *
> [https://github.com/apache/kudu/blob/89e2715faf96afe0b67482166fda9c8699e8052f/docs/prior_release_notes.adoc?plain=1#L143-L145]
> * https://issues.apache.org/jira/browse/KUDU-3375
>
> However, with this commit
> ([https://github.com/apache/kudu/commit/e65ea38a4860c007d93ada9c991bccec903a80b1)]
> , string gauge related to clock_ntp_status is exposed.
>
> {code:java}
> # HELP kudu_master_clock_ntp_status Output of ntp_adjtime()/ntp_gettime()
> kernel API call
> # TYPE kudu_master_clock_ntp_status gauge
> kudu_master_clock_ntp_status{unit_type="state"} now:1706665936956760
> maxerror:70013 status:ok {code}
> It prevents prometheus operators from collecting prometheus metrics for kudu.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
