Re: Some feature notes for monit

[Vlada Macek]

Martin Pala wrote:

>> ### Access time changed by checks
>
> This is not checked, possibly it could be. For performance reason i
> personaly usualy turn off atime on filesystem, thus this is not my
> cup of tea - maybe other developers are interested in it, or you can
> give it a stab ;) In such case, probably the timestamp test syntax
> should be more general ...

I meant for example file checksum test which is changing access time.
Upon startup, monit could check whether there is atime feature turned
on/off on the filesystem, mark to the service structure and later do
restoring atime for files or not.


>> ### Missing check: general lstat(2) variable check
>
> Described 'lstat change' test could be usefull only for definitely
> static objects, which is special case => because of general usage, it
> is probably better to test for specific attributes of monitored
> object, which are important to monitor, with appropriate limits
> definition. If there are some tests for specific attributes missing,
> we can add them.

When I monitor configuration files and binaries of their daemons, I
would use my 'lstat change' test. Such files are definitely static
objects from my point of view and I wish to detect any tampering with
them. I expect storing the lstat structure to the monit.state file so
restarting monit does not have impact to the variable test.

Furthermore the line `if changed stat then alert' is brief and makes
monitrc a better readable. :-)

VM

signature.asc
Description: OpenPGP digital signature