Eve.json filename options


I have seen that the eve.json filename can be formatted this way filename: eve-%Y-%m-%d-%H:%M:%S.json but I would be interested on a no-time-dependant identification. Something like eve-1.json for the first log, and eve-2.json for the second one, and so on, increasing each time a new log file is created. The purpose of this is having the log files in order without time consideration. Ideally it would be persistent, I mean, when suricata starts would check for past logs in the directory and continue the count from the biggest one.

I have been checking and didn`t find something like this but just to make sure: There is nothing like this, right? Or something I could achieve something similar with?

I also have been checking the code to see if I would be able to develop it myself, I am still trying to understand the code so I might be really wrong but to achieve what I want it would be something inside

SCLogFilenameFromPattern function or maybe an alternative function to this. Am I in the correct path?

By the way, I have been checking the 6.0.13 code, where did this go on the 7.0.x version?

Hi, it seems to me that eve.1.json etc already happen if multi threaded eve output is enabled.

I believe @Jeff_Lucovsky is the most well versed in this part of the code. Let’s see if he has some insights to offer

If I understand correctly, according to the documentation ( threaded-file-output , the files do get the name “eve.x.json” but the problem is that when HUP signal is sent the name is still the same, and I want it to increase x +1. This is because a log transporting issue that I am having, I want files to be in order(that’s why the increasing number) and not be named the same(for avoiding duplicated sending) , but time is probably inconsistent y my system.

Reading my post, I think I didnt explained it correctly so I am going to edit it to improve it.

Are you using logrotate in your deployment?

If not, I suggest that you consider using it to help with custom names.

The numeral in the treaded output filenames has not meaning outside of Suricata internals and is thus, cannot be handled that way.

I am currently not using logrotate because it changes the name of the original file when rotated. So when this was done, it was treated as a new file and it was transported again, creating duplicated logs. What I need is to directly create the file with a new name.
Then I guess I might have to develop it myself, is this SCLogFilenameFromPattern the correct way to start from?

I’m not sure what your use case is but I would advise using logrotate with a different workflow, if needed.

You’d have to experiment a bit to see; the threaded (high performance) case wouldn’t be handled with mods to that function.

You can get a close approximation to this. Assuming you don’t have the threaded option enabled (which is not enabled by default), set the filename to something like:

      filename: eve-%s.json

this will use the UNIX time in seconds. On each HUP it will get a new timestamp. Your custom transporter/rotator may need some logic tho, like:

  • find the most recent log file
  • send sighup
  • transfer and delete most recent log file before the hup

Starting and 1 and working forward complicates things when the above should work just as well.

Yes, I have something really similar to this right now. I just wanted to get rid off time dependency to order the files.