Some clients disappear from web interface but are still present in databse

[Wayne Workman]

@Claude-Girard said:

@Matthieu-Jacquart said:

I updated and restore database from last save (last week-end), and 1 hour later at least 3 hosts disappear, on 3 different groups, I made no changes…

edit : we forgive you, you all make great job

I can confirm this bad news.
I wanted to control my database, and I made this request in phpmyadmin:
SELECT * FROM hosts WHERE hostID NOT IN (SELECT hmhostID FROM hostMAC)
It returned 93 results
10 minutes later, the same request returned 101 results.
No operation made on fog interface between the requests.

Do I need to upgrade client too ? I don’t think so, some disappeared computers have legacy client installed

That returned 41 rows for me… and I’m using the legacy client…

I’m going to restore a previous snapshot and see if that number decreases.

Yes, I’ve confirmed that the snapshot I just restored - when I ran that command, i got an empty set returned… ouch.

I’m not updating till we know this is fixed.

However I’m happy to update as much as the @Developers would like at home.

[Trevelyan]

Oops, just seen this - I guess I was searching for the wrong things - although I ended up doing everything else mentioned here…

I’ll watch with intrigue…

[Joe Schmitt]

@Claude-Girard @Matthieu-Jacquart @Wayne-Workman @Trevelyan @Brian-David
I need the fog.log file from disappeared hosts (preferably relatively soon after they vanished). Logs from the new client will be most helpful.

[Claude Girard]

How do I send it ? MP ?

[Wayne Workman]

@Claude-Girard said:

How do I send it ? MP ?

Just upload it to the forums, @Jbob can get it that way.

[Matthieu Jacquart]

Ok, I’m making a test : I empty database from all hosts and everything associated to hosts.
After that, I import all hosts with a csv file, and I’m putting them in their groups to modify parameters.
I started around 1 hour, for the moment all my hosts are still present…

[Claude Girard]

@Matthieu-Jacquart said:

I started around 1 hour, for the moment all my hosts are still present…

After restoring backup:
I deleted groups, and recreate them, include hosts in groups.
After 2 hours, hosts are still present.
But now some hosts dissapear from some groups.

[Tom Elliott]

@Claude-Girard So what you’re saying is, everything was fine for nearly 2 hours, then they just started removing themselves? Do you have multiple nodes? If you do have multiples, are they all on the same revision?

[Claude Girard]

@Tom-Elliott said:

@Claude-Girard So what you’re saying is, everything was fine for nearly 2 hours, then they just started removing themselves? Do you have multiple nodes? If you do have multiples, are they all on the same revision?

No, hosts are presents and stay presents, but some groups are modified, hosts disappear from group members only

[Tom Elliott]

@Claude-Girard WHEW

Thank you,

While that is an issue on it’s own, at least the host’s (themselves) are still in tact.

[Tom Elliott]

What are the steps I can take to attempt reproducing hosts losing their group associations?

[Claude Girard]

@Tom-Elliott said:

@Claude-Girard WHEW

Thank you,

While that is an issue on it’s own, at least the host’s (themselves) are still in tact.

Yes that’s the most important
No time today but tomorrow I’ll give more infos about group issue.
I’ll do more tests because after my restore, I prefered clean database by deleting some hosts and recreating them.
The ones that leave group were not.
I’ll try to delete them and see if after recreate problem pesist.

[Claude Girard]

@Claude-Girard And good news:
My sql request:
SELECT * FROM hosts WHERE hostID NOT IN (SELECT hmhostID FROM hostMAC)
hours ago !!!
returns 0 rows since last fog update,

[Matthieu Jacquart]

24 hours later, my 312 hosts are still in database and in web interface

[Tom Elliott]

I did make a commit earlier did reinsert the same type of problem. When informed, I reverted the code. I have re-reverted the code after testing many potential opportunities of the random delete issue. i’m not 100% it’s good to say, yay or nay, but from the limited potential I could test within, all seems to be normal.

This I have tried to “see it”

Add hosts to a group, with many updates at random times and settings.

Falsely having the client run tasks.

Properly having a client run tasks.

Updating the host directly (as well as others).

From different pages, insert the host or hosts as seen fit.

I’m not even seeing modules randomly appear or disappear.

Hopefully by the time i get in to work tomorrow I can put a solve setting on this thread.

[Claude Girard]

Hopefully by the time i get in to work tomorrow I can put a solve setting on this thread.

Yes I think you can.
Hosts are still here !!!
I can’t reproduce group issue, during 1 hour I delete hosts from groups, include them in existing or new groups, modify by group image association, service settings, active directory, and snapins.
I lost 1 member in a group only one time, 66 members at start, and 65 after modifying settings in the group, not sure but I think after active directory settings.

I decided to do that because I thought that after this bug my database was in an inconsistancy state.

I sent these 3 requests to my database:
SELECT * FROM hosts WHERE hostID NOT IN (SELECT hmhostID FROM hostMAC)
Before update, I had a lot of rows in result, and growing. But now it’s ok.

SELECT * FROM hostMAC WHERE hmhostID NOT IN (SELECT hostID FROM hosts)
Never had result, was ok and is still ok

SELECT * FROM snapinAssoc WHERE saHostID NOT IN (SELECT hostID FROM hosts)
Returned a few rows in result, maybe not directly due to this bug, maybe this table was corrupted before.

I think that tables refer to hosts by Mac adresses, except snapinAssoc table. Is it right ?
So with these 3 requests, I can find bad hosts and bad snapin assoc, and I delete them from database.
But maybe I forgot one or several tables to check ?

Thank you for you job !

[Brian David]

Alright, I’ve currently got git version 4682. I started from a fresh database and imported my hosts via a csv file, then I adjusted all my service settings. It’s been going for about two hours now, and all the hosts are still in the web interface. Also, I have yet to see any settings change in unpredictable ways. So far so good!

[Tom Elliott]

Then I’m going to solve the thread. As far as I can tell, all of the sources of reports of this horrendous problem are no more occurring on current trunk.

Thank you all for your patience with this.