Bug #6118: cf-agent execution is killed if cf-execd is restarted (for example on a 5a.m daily restart) - Rudder - Issue Tracker

Custom queries

8.1 beta
Bugs triage
CS - by priority
CS - to review
My TR
Need technical review
Open bugs (all)
Open bugs (UI - UX)
Open UX defects (all)
Opened by users

Actions

Copy link

Bug #6118

closed

cf-agent execution is killed if cf-execd is restarted (for example on a 5a.m daily restart)

Added by Nicolas CHARLES almost 10 years ago. Updated about 7 years ago.

Status:

Rejected

Priority:

Assignee:

Nicolas CHARLES

Category:

System techniques

Target version:

4.3.0~beta1

Pull Request:

Severity:

Critical - prevents main use of Rudder | no workaround | data loss | security

UX impact:

User visibility:

Operational - other Techniques | Technique editor | Rudder settings

Effort required:

Priority:

Name check:

Fix check:

Regression:

Description

We have a promise to restart services at 5am every day. Unfortunately, it does not behave as expected as stopping executor daemon kills apparently the agent, causing it to skip a run (it does show especially with advanced reporting plugin)

In 2.10, it merely truncate the logs in outputs folder, but in 2.11 and 3.0 it skips the runs completely

Related issues 1 (0 open — 1 closed)

Related to Rudder - Bug #7274: The daily cf-execd and cf-serverd restart should use SRC on AIX

Released

Benoît PECCATTE

2015-10-13

Actions

Issue # Delay: days Cancel

History
Notes
Property changes

Actions

Copy link

Updated by Vincent MEMBRÉ almost 10 years ago

Target version changed from 2.11.6 to 2.11.7

Actions

Copy link

Updated by Vincent MEMBRÉ almost 10 years ago

Target version changed from 2.11.7 to 2.11.8

Actions

Copy link

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.8 to 2.11.9

Actions

Copy link

Updated by François ARMAND over 9 years ago

Status changed from New to Discussion
Assignee set to Nicolas CHARLES

Just to be sur I understand the bug here: the problem is not the 5am stop, it's that if we restart cf-execd, then cf-agent is killed (and so, if a run was on going, it is killed). Is it that ?

Actions

Copy link

Updated by Nicolas CHARLES over 9 years ago

Assignee changed from Nicolas CHARLES to François ARMAND

The problem from the user point of view is that every day at 5 am there is a dent in reporting, with no answer everywhere

The cause is that the agent kills cf-execd every day at 5 am (it's part of the promises), and in result it kill the agent that was launched by cf-execd, interrupting the run.

Probably killing cf-execd should not kill cf-agent, but i'm not sure about this one.

Actions

Copy link

Updated by François ARMAND over 9 years ago

Subject changed from cf-agent execution is killed every day at 5am to cf-agent execution is killed if cf-execd is restarted (for example on a 5a.m daily restart)

I think they should be independant, but don't quite see the implication here.

Well, in all cases, I only see two possibilities:

- either cf-execd and cf-agent are independant, and killing the parent does not kill the child;
- or they are not (and we don't want them to be), and we need to have a gracefull restart for cf-execd. That may be extremelly tricky, because deciding between a stalled cf-agent and one that is just taking a lot of time to finish its tasks may not have a definitive answer...

I think that cf-execd manages a lot of thing (sending emails, getting stats of execution, etc) so perhaps we won't be able to split them, even if we wanted.

For now, I don't see other option. Perhaps we should just not kill the process every days, and consider a kill for what it is - an interruption.

Actions

Copy link

Updated by Nicolas CHARLES over 9 years ago

The issue here is not in case of stalled agent, it's really that we are preventively restarting cf-execd every day, and it kills the agent.

Restarting the executor daemon used to be part of defaut CFEngine masterfiles, but now they are not anymore, (at least since 08/2013)
So maybe we should remove this preventive restart

Actions

Copy link

Updated by François ARMAND over 9 years ago

Nicolas CHARLES wrote:

The issue here is not in case of stalled agent, it's really that we are preventively restarting cf-execd every day, and it kills the agent.

I understand that, what let think that I was talking about stalled agent for the restart ? With graceful restart, you have basically 3 cases to consider:

- "please stop" : [ok I'm stopped]
- "please stop" : [nothing happen : stall or just finishing ?] ;
- ok, after some times / some retries, it's stopped;
- even after some times / some retries, it is not stopped => stalled or not ? When to kill ?

Restarting the executor daemon used to be part of defaut CFEngine masterfiles, but now they are not anymore, (at least since 08/2013)
So maybe we should remove this preventive restart

That could be the workaround for that.

Actions

Copy link

Updated by François ARMAND over 9 years ago

Assignee changed from François ARMAND to Nicolas CHARLES
Reproduced set to No

Actions

Copy link

#10

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.9 to 2.11.10

Actions

Copy link

#11

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.10 to 2.11.11

Actions

Copy link

#12

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.11 to 2.11.12

Actions

Copy link

#13

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.12 to 2.11.13

Actions

Copy link

#14

Updated by Vincent MEMBRÉ over 9 years ago

Target version changed from 2.11.13 to 2.11.14

Actions

Copy link

#15

Updated by Vincent MEMBRÉ about 9 years ago

Target version changed from 2.11.14 to 2.11.15

Actions

Copy link

#16

Updated by Nicolas CHARLES about 9 years ago

Related to Bug #7274: The daily cf-execd and cf-serverd restart should use SRC on AIX added

Actions

Copy link

#17

Updated by Vincent MEMBRÉ about 9 years ago

Target version changed from 2.11.15 to 2.11.16

Actions

Copy link

#18

Updated by Vincent MEMBRÉ about 9 years ago

Target version changed from 2.11.16 to 2.11.17

Actions

Copy link

#19

Updated by Vincent MEMBRÉ almost 9 years ago

Target version changed from 2.11.17 to 2.11.18

Actions

Copy link

#20

Updated by Vincent MEMBRÉ almost 9 years ago

Target version changed from 2.11.18 to 2.11.19

Actions

Copy link

#21

Updated by Vincent MEMBRÉ over 8 years ago

Target version changed from 2.11.19 to 2.11.20

Actions

Copy link

#22

Updated by Vincent MEMBRÉ over 8 years ago

Target version changed from 2.11.20 to 2.11.21

Actions

Copy link

#23

Updated by Vincent MEMBRÉ over 8 years ago

Target version changed from 2.11.21 to 2.11.22

Actions

Copy link

#24

Updated by Vincent MEMBRÉ over 8 years ago

Target version changed from 2.11.22 to 2.11.23

Actions

Copy link

#25

Updated by Vincent MEMBRÉ over 8 years ago

Target version changed from 2.11.23 to 2.11.24

Actions

Copy link

#26

Updated by Vincent MEMBRÉ about 8 years ago

Target version changed from 2.11.24 to 308

Actions

Copy link

#27

Updated by Vincent MEMBRÉ about 8 years ago

Target version changed from 308 to 3.1.14

Actions

Copy link

#28

Updated by Vincent MEMBRÉ about 8 years ago

Target version changed from 3.1.14 to 3.1.15

Actions

Copy link

#29

Updated by Vincent MEMBRÉ about 8 years ago

Target version changed from 3.1.15 to 3.1.16

Actions

Copy link

#30

Updated by Vincent MEMBRÉ about 8 years ago

Target version changed from 3.1.16 to 3.1.17

Actions

Copy link

#31

Updated by Vincent MEMBRÉ almost 8 years ago

Target version changed from 3.1.17 to 3.1.18

Actions

Copy link

#32

Updated by Vincent MEMBRÉ almost 8 years ago

Target version changed from 3.1.18 to 3.1.19

Actions

Copy link

#33

Updated by François ARMAND over 7 years ago

Severity set to Critical - prevents main use of Rudder | no workaround | data loss | security
User visibility set to Operational - other Techniques | Technique editor | Rudder settings
Priority set to 0

I'm setting it to critical because there is no workaround, and it breaks the main purpose of rudder. On the other hand, I'm setting it to operationnal because the problem is not supposed to happen on tests.

Actions

Copy link

#34