VMware Cloud Community
linux-guru
Contributor
Contributor

FTP-Service doesn't close connections after testing

Hi,

I am monitoring my FTP-Daemon (proftpd) by using the "default" FTP service of hyperic.
Login-settings are correct, graphs and alerts work.
Unfortunately hyperic doesn't close the connection after doing its check which results now (after 18 hours of having the agent running) in about 50 open connections.

Are there any tweaks / workarounds available or could this problem be solved asap?

Cheers

Tobias
Reply
0 Kudos
5 Replies
john_hyperic
Hot Shot
Hot Shot

Go ahead and log a bug for this in JIRA (http://jira.hyperic.com/) and
I'll see if we can get a plugin fix to you ASAP.

John Sachs
Hyperic Support

Reply
0 Kudos
dougm_hyperic
VMware Employee
VMware Employee

I installed proftpd on a CentOS VM and my HQ agent has been
monitoring the ftp port for the past ~36 hours with no leaked
connections:

% sudo netstat -atp | grep ftp
tcp 0 0 *:ftp
*:* LISTEN 26400/proftpd: (acc

% /usr/sbin/proftpd --version
- ProFTPD Version 1.2.10

Can you post the output of the two commands above run on your system?


Reply
0 Kudos
john_hyperic
Hot Shot
Hot Shot

I also could not reproduce the problem using vsftpd 1.2.1 on Fedora Core 2.

Reply
0 Kudos
linux-guru
Contributor
Contributor

Seems that there was something different wrong.
Because of the fact that auto-discovery of MySQL-Tables on about 5 servers with about 200 Databases each made HQ doing very strange things and also made the server where HQ-server ran, nearly unusable.
IMHO there seems to be too complex checking of each DB/table which brings the server (P4 1,6 GHz, 1GB RAM) nearly down.
After doing a complete re-install of the server and all clients, the problem with FTP connects went away.
Another thing is saw, was that HQ seems to open many connections to the database-server when monitoring databases and tables. Running the whole thing produced an error of MySQL which stated that there were too many connections.
Maybe this is another thing which happens when the server gets overloaded OR it is a misfunction of HQ. I don't know, but it might be an issue the HQ-developers can have a look to.

Cheers

Tobias
Reply
0 Kudos
admin
Immortal
Immortal

Hi Tobias,

There is an open bug for the mysql plugin to add better caching of metrics. Many of the metrics we gather come from the same place, so we should be able to get them all at once. Currently the plugin issues a query for each metric value. This results in many queries on installations that have many databases or tables. The bug report is at:

http://jira.hyperic.com/browse/HHQ-29

A workaround for the problem would be to manually create the MySQL server for the databases you are interested in, rather than letting the MySQL Process server detect them all.

As for the amount of data that is being collected, you can modify the default set of metrics and their intervals by going to Administration->Monitoring Defaults Configuration and then clicking on the 'Edit Metric Template' button for the MySQL server type you are using. That way HQ is only collecting the metrics you are interested in.

Hope that helps,
-Ryan
Reply
0 Kudos