Adding Varnish HTTP Cache input plugin #1173
Conversation
| ## By default, telegraf gather stats for 3 metric points. | ||
| ## Setting stats will override the defaults (["%s"]). | ||
| ## stats may also be set to ["all"], which will collect all stats | ||
| stats = ["%[1]s"] |
There was a problem hiding this comment.
I've used formatted string to avoid repeating the defaultStats list in multiple places. I'll clean it up and update the README and CHANGELOG as per your comments on my other PRs.
| cmd := exec.Command(cmdName, cmdArgs...) | ||
| var out bytes.Buffer | ||
| cmd.Stdout = &out | ||
| err := cmd.Run() |
There was a problem hiding this comment.
use internal.RunTimeout here (https://github.com/influxdata/telegraf/blob/master/internal/internal.go#L166)
robinpercy-xm
force-pushed
the
pr/input-varnish
branch
from
d074e9e to
895987f
Compare
- Linked to varnish in input README - Updated README/CHANGELOG - Cleaned up sampleConfig to remove formatting - Shorted lines to 80 chars (except where test input requires long strings) - Using internal.RunTimeout to wrap call to varnishtat - Added dummy file for windows
robinpercy-xm
force-pushed
the
pr/input-varnish
branch
from
895987f to
d3646df
Compare
|
what would you think about the first part of the measurement being lower-cased & integrated into the measurement name? ie, instead of it could be |
|
although what you have is the format that varnish gives it in....so it might be best to keep it as-is |
|
Yeah, I'm inclined to leave them as they are. As you said, it mirrors what varnishstat provides, which makes query building intuitive. The InfluxDB docs seem to recommend tags over measurement names for organizing. Though, the stated advantages don't really apply to this scenario, since the field names are all unique across sections anyway. Series cardinality stays the same in either case. Are there any other considerations around transmission or storage that could tip the scales? |
collects metrics from the varnishstat command