Upgrade 1.2.19 to 1.2.21 ERROR PHP ERROR: Uncaught Error: Call to undefined function boost_debug()

[nicolatron]

Hello,

Just upgraded from 1.2.19 to 1.2.21 (cacti and spine installeed from the version on download page), I have a setup with one main cacti and 2 collectors.

I'm getting this on the main poller log every poller cycle:

2022-06-07 08:45:07 - CMDPHP PHP ERROR Backtrace: (CactiShutdownHandler())
2022-06-07 08:45:07 - ERROR PHP ERROR: Uncaught Error: Call to undefined function boost_debug() in /opt/cacti/lib/boost.php:1257 Stack trace: #0 /opt/cacti/poller.php(811): boost_poller_bottom() #1 {main} thrown in file: /opt/cacti/lib/boost.php on line: 1257
2022-06-07 08:45:07 - SNMPAGENT WARNING: No notification receivers configured for event: cactiNotifyDeviceFailedPoll (CACTI-MIB), severity: medium
2022-06-07 08:45:07 - SYSTEM STATS: Time:3.8074 Method:spine Processes:1 Threads:1 Hosts:13 HostsPerProcess:13 DataSources:565 RRDsProcessed:0

Searched for similar errors but found nothing so far.
Any recommendations?.

But at least the polling on the collectors seems to go fine, couse i'm getting new data on the graphs.

Graph displaying seems to be very slow (not to the point where it's unbearable, but maybe it takes half a second to load one graph, where it retrieved maybe 10 on 1 second before the upgrading to 1.2.19), weathermap plugin seems not to be working (used latest git version from phalek), but I see no message related to weathermap on the log.

Tried to upgrade and restar whole servers (to SLES 15 SP2). But behaviour didn't change.

Also executed this on databases (all 3 of them):

MariaDB [cacti]> ALTER TABLE poller_output_boost ENGINE=memory;
Query OK, 0 rows affected (0,022 sec)
Records: 0 Duplicates: 0 Warnings: 0

MariaDB [cacti]> ALTER TABLE poller_output ENGINE=memory;
Query OK, 0 rows affected (0,007 sec)
Records: 0 Duplicates: 0 Warnings: 0

MariaDB [cacti]> ALTER TABLE poller_output
-> MODIFY COLUMN output varchar(512);
Query OK, 0 rows affected (0,006 sec)
Records: 0 Duplicates: 0 Warnings: 0

MariaDB [cacti]> ALTER TABLE poller_output_boost
-> MODIFY COLUMN output varchar(512);
Query OK, 0 rows affected (0,007 sec)
Records: 0 Duplicates: 0 Warnings: 0

I have very long ouput becouse of F5 ridiculously long table indexes.

Not related to this, but before I upgraded I was having issues with some graphs not working, (I think because I imported some template that overwrite my most used interface traffic bit/sec 64-bit counter graph template and/or related datasources, and so rrd files information on max values did not match the one on the datasources, after running datasource troubleshooting recommendation was to rrdtune the file. So I thought... if I'll end up regeneration some hundred graphs anyway might as well upgrade and recheck everything then.

[nicolatron]

I've rolled back to 1.2.19.

[TheWitness]

Fixed here, though I use boost heavily and have never received this error. Maybe it has something to do with the fact that I've never used the 'debug' feature in boost. Sorry about that.

https://github.com/Cacti/cacti/issues/4810

[nicolatron]

Thanks TheWitnees, was not aware that I had boost debug activated (or where can it be activated). I'll look for it, just in case I also have it active in my production machine.
I'm planning on taking off one of the 2 collector (delete collector from main).
Change it's include/config.php file commenting out the $rdatabase_* entries and see if it acts a standalone cacti (with the same database production machines have more or less?). Then Upgrade to 1.2.21 there and test again (a lot of ifs).
I guess I'll get 1.2.x branch from git instead of the tarball (https://github.com/Cacti/cacti/tree/1.2.x).
BTW is there any documentation on how to promote a collector to main?. I think that'll be handy to make a test machine for upgrades and the like.

[TheWitness]

It's not really supported. If that remote server will only poll it's own devices, what you should do instead is:

1) On main server, make a completed MySQL dump
2) Disable the remote poller to prevent it from performing additional updates
2) Restore that database on the remote server
3) RSYNC the RRDfiles
3) Login to that server and remote the main server devices leaving only the remote server devices
4) Promote the remote server devices to main collector devices
5) Remove the former remote poller from Data Collectors
6) Re-enable polling
7) Do an RRDfile cleanup using the Utility provided

If it's going to poll everything, you can cut a number of steps out of the process above. I would think that Sean Mancini might want to do a video on this. You might want to create a Feature Request on GitHub too.

[nicolatron]

Another thing to remember, remove or comment the RDB_* entries on spine.conf on the collector to be removed (and promoted). If using spine.