spine 1.2.9 regularly crashing

[hektor]

I see spine 1.2.9 crashing quite frequently on both of my pollers:

----------------------------------------------------
*** Error in `/usr/local/spine/bin/spine': double free or corruption (out): 0x00007f45d8007540 ***
======= Backtrace: =========
/lib64/libc.so.6(+0x81679)[0x7f4615af3679]
/usr/lib64/mysql/libmysqlclient.so.18(+0x256d5)[0x7f4616c006d5]
/usr/lib64/mysql/libmysqlclient.so.18(+0x2a3d6)[0x7f4616c053d6]
/usr/lib64/mysql/libmysqlclient.so.18(+0x27b84)[0x7f4616c02b84]
/usr/lib64/mysql/libmysqlclient.so.18(mysql_close+0x1c)[0x7f4616c02bbc]
/usr/local/spine/bin/spine[0x40cf3d]
/usr/local/spine/bin/spine[0x40f2ac]
/lib64/libpthread.so.0(+0x7e65)[0x7f461634de65]
/lib64/libc.so.6(clone+0x6d)[0x7f4615b7088d]
----------------------------------------------------

Can I contribute to resolving that?

System is CentOS 7, spine is selfcompiled.

[Osiris]

Grab the latest 1.2.x branch and report back.

[hektor]

Does not look better in 1.2.10

[TheWitness]

1.2.x branch, not 1.2.10.

[hektor]

cacti itself as well as spine?

[TheWitness]

If you choose too. We are basically in a cool down phase right now. Just received another update on spine yesterday that should quell al the issues. Spine has been pretty reliable for years, it's when we touch it that things break for a few releases.

[hektor]

I tried and tried but installing (or upgrading) is always a mess.

At the moment I get stuck on {"Step":"97","Eula":"1"} or {"Step":"98","Eula":"1"} with the page reading "we are ready".

Numerous tries to re-install.

[hektor]

7th or so install now worked out.

It's a remote poller ( I have another one, but this is not available for experiments)

Spine seems to be stable, but has not so much to do yet
but I have this on the remote poller:

2020/04/02 17:38:05 - PCOMMAND Device[4483] WARNING: Recache Event Detected for Device
2020/04/02 17:38:05 - ERROR PHP ERROR: Call to undefined function poller_push_reindex_data_to_poller() in file: /usr/local/cacti/lib/data_query.php on line: 385

[hektor]

Funny: that get written over after I edited it to "poller_push_reindex_data_to_main".

Does it get overwritten from the main poller?

[hektor]

OK it does.

Was the change in lib/data_query.php, line 385 correct?

[TheWitness]

If you mean lib/poller.php. We moved it to global.php.

[hektor]

???

IMHO there is a bug in lib/data_query.php, line 385

[hektor]

Alright, opened an issue.

[hektor]

Looks like I have found out something:

the crashes seem to correspond to bad synchronisation of system clocks of the various pollers.

[hektor]

And as I wrote it, I got another one.