topic I think that builtin:host.disk.usedPct metric is wrong because the total space metric is wrong. in Alerting

I think that builtin:host.disk.usedPct metric is wrong because the total space metric is wrong.

itamarmesa — Fri, 15 Mar 2024 11:39:37 GMT

Seeing directly the sizes of a mounted point of a disk in my host, it gives me some stats:

84.17% used disk space,

2,412383548915386 TiB disk used (2652443762688 bytes / 1024 / 1024 / 1024 /1024)

2,86578132212162 TiB disk used (3150959886336 bytes / 1024 / 1024 / 1024 /1024)

While in Dynatrace, the same disk using the builtin metric:

(builtin:host.disk.usedPct:filter(and(or(in("dt.entity.disk",entitySelector("type(disk),not(tag(~"WarningDisk:90~"))")))),series(avg,and(ge(85),le(90)))):splitBy("dt.entity.disk","dt.entity.host"):parents):filter(eq("dt.entity.disk",DiskA),eq("dt.entity.host",HostA))

gives me a 85.43%, which is wrong.

Also, seeing the metrics in the host makes everything clear:

We can see that the pct calculation fails probably because the total space is wrong, because 2,41... / 2,82... TiB ~ 85,43%.

Why is this? Is it that the agent wrong? Is it that the total space calculation is wrong?

Thanks.

Re: I think that builtin:host.disk.usedPct metric is wrong because the total space metric is wrong.

ChadTurner — Wed, 27 Mar 2024 12:34:13 GMT

It looks like the metric that you are getting is set to be an average could this be the discrepancy? Have you changed the aggregations to see if the data set comes back more inline with what you are expecting?

Re: I think that builtin:host.disk.usedPct metric is wrong because the total space metric is wrong.

itamarmesa — Wed, 27 Mar 2024 14:29:20 GMT

Actually I found out what was the discrepancy, I just didn't share it here.

For Dynatrace, the amount of Disk total is the disk available + disk occupied, which is actually not the total, but the total disk usable, since It doesn't take into account the non available part of the disk. I mean, It is okay now that I know, but since Zabbix's default percentage for disks would take into consideration the actual total... I just wanted to know if there was some failure in Dynatrace or something.