Remove deprecated features: compatlog, commandpipe

[dnsmichi]

#6031
#6033
#6032

[rippiedoos]

If you remove this feature, could you please remove it properly (or fix it in 2.10). Version 2.10 breaks this feature but it's still working in 2.9.2. It is even in the documentation. If you enable the statusdata-feature in 2.10 the objects.cache-file isn't created anymore.
If you run the command icinga2 object list --type=StatusDataWriter in 2.10 it crashes. In 2.9.2 it returns the proper output.

[dnsmichi]

The above was fixed, on removal it won't influence anything else.

[Al2Klimov]

@lippserd Livestatus which is used by one of our key customers uses the compat logger for historical tables. Shall we care about this?

[Al2Klimov]

CC @chrnie

[Al2Klimov]

To be precise: It uses the logs produced by the compat logger.

[dnsmichi]

@lippserd Livestatus which is used by one of our key customers uses the compat logger for historical tables. Shall we care about this?

We cannot remove that for now, that's why this issue is subject for removal from 2.13.

[chrnie]

compat logger is an important feature for thruk's logfile cache.
https://www.thruk.org/documentation/logfile-cache.html
Finally, the entries are to end up in a database.
If this feature is deleted, it would be nice if there was another possibility, apart from importing the IDO. Since not using the IDO was the initial reason to use livestatus.

[Al2Klimov]

You're intentionally not using the IDO? Why?

And yes, there will be another possibility in v2.12 – IcingaDB.

[Al2Klimov]

Yes, we did.

[johannesst]

Hello everybody,

I have a question concerning the planned removal of commandpipe. Will it be removed with the other deprecated features or might it be removed non the less even when the other removals are stalled?

I'm asking since we are now migrating from Nagios to Icinga2 and our IBM host administrators use NSCA for passive host checks. Although they will migrate to NRPE at some point at the moment there is no other possibility.

We don't want to get a situation where we would have to use the last working Icinga2 version with commandpipe thus not getting security features anymore. So: Is there a possible date when this removal will take place?

Best Regards, Johannes

[Al2Klimov]

statusdata and checkresultfilereader can be removed.

Did we already deprecate the command pipe?

@lippserd Shall we throw those away in 2.14?

[Al2Klimov]

v2.9

• Deprecate CheckResultReader #6063
• Deprecate Statusdatawriter #6198

v2.11

• Deprecate ExternalCommandListener feature ('command') and adjust log warnings to the roadmap #7078

I'd be nice and only throw out the v2.9 ones in v2.14.

[Al2Klimov]

Note

We can't drop the command pipe until we have a comparable performant mechanism for feeding Icinga with bulk check results. (@Thomas-Gelf)

[Al2Klimov]

Idea regarding command pipe (for one nice day): Remove not all at once, but throw out all not mission critical functionality of the command pipe. (But not in v2.14.)

[Al2Klimov]

FWIW, it seems (#10397) Livestatus can process check results on its own, so doesn't need command pipe. CC @chrnie

[Al2Klimov]

@chrnie just confirmed that Thruk, and hence our fellow customer too, does NOT use the command pipe. The customer is also preparing to deal with absence of compatlog.

@lippserd I suggest to remove the command pipe in our next .0 version. Like, fellow users, you got the shiny new

• Sync dependencies to Redis #10290

so, conversely, give up the command pipe please.

Yes, I know, we have now fancy topic releases etc. But I can hardly imagine how this issue fits into the topic releases pattern. And, as legacy stuff, the command pipe is a bit of a maintenance burden. Examples:

• Checkable#ProcessCheckResult(): discard🗑️ CR or delay its producers shutdown #10397
• Don't fail to remove obsolete downtimes #10062
• Downtime::AddDowntime(): NULL-check pointer before deref not to crash #10049
• Checkable#ProcessCheckResult(): only clean up ack comments older than check result #9718
• Icinga DB: correct ack comments' is_sticky #9287
• API Action "remove-downtime": Also remove child downtimes #8913
• Feature: Replace std/boost::bind() with lambdas #8601
• Fix null pointer dereferences when deleting objects while scheduling downtimes #8588
• Drop passive check results for unreachable hosts/services #8267
• API: Display a correct status when removing a downtime #8104
• Use std::mutex, not boost::mutex #8099
• Refuse acking an already acked checkable #7695
• IcingaDB: populate icinga:history:stream:acknowledgement #7667

[yhabteab]

CheckResultReader and StatusDataWriter are already removed with v2.14. Only the two are remaining now but won't drop them with v2.15 yet.

[Al2Klimov]

🤔

Not the most obvious decision not to remove any of them in the upcoming .0, given we removed even two in the last .0 version.