I am sure I read something about IIS and log file modtime in here somewhere, but I can't find that post again.
Anyway, from the Splunk docs and other comments, I gather that IIS not updating modtime is the reason why I have odd issues with IIS logs.
Can anyone confirm this?
What happens is that suddenly, some time during the day, my IIS logs stop coming in. I haven't sat around to watch it for hours, but I can see that next morning, all the data is there.
I haven't confirmed, but I would guess that the log file switch at midnight is what triggers a read of the file, and the the data is indexed.
The question then is: is there a solution to that problem? Can I make Splunk follow the log continuously (without duplicate indexing results, of course)?
As said, I thought I read the answer to that question in here before, but I can't find it again.
One odd thing about IIS logs I noted is that they, by default, rotate every 24hours at 0:00 GMT (+000), in other words, they actually rotate in the middle of the day over here (depending on your regional/timezone settings).
To minimize any problem I changed the IIS to rotate logs every hour... it not only gives me smaller files to work but also more predictability. I also added the line to my inputs.conf:
ignoreOlderThan = 7d
so it won't re-scan older files. After 8 days I have a PS script do remove/archive the logs as well.
Not related, but I also suggest installing the Advanced Logging module for IIS... gives you the ability to get more information and configure custom fields in your http log.
That's a config choice in IIS: you can select whether to use local time for file name and rollover. If you don't, it will use UTC. For companies with many log sources across timezones, aligning everything to UTC can be beneficial.
I'll check out the Advanced logging module, thanks for the tip!