Since this weekend I suddenly have a bunch of hosts that don't exist. A script that is meant to alert if any host has not sent logs in 24 hours started reporting hosts named "recover-padding-1" through "recover-padding-84" as not checking in. When I look at the Summary they are also listed but no event entries are listed.
How and why did 84, "recover-padding-#" hosts, get added to my host list with out any events? Why do I still get output on my script after I "deleted" them? Usually when I "delete" something it changes the date on a "host" last contact to be the date when Splunk started collecting data.
Query Terms: '| metadata index=main type=hosts | eval age = now()-lastTime | where (age > (1*86400)) AND (age < (7*86400)) | sort age d | convert ctime(lastTime) | fields age,host,lastTime'
Link to results: arthamm_search_bWlub3JfaG9zdF9ub3RfcmVwb3J0aW5n_at_1292864400_1115244169">http://nascsplunk:8000/app/search/@go?sid=scheduler_arthamm_search_bWlub3JfaG9zdF9ub3RfcmVwb3J0aW5n_at_1292864400_1115244169
Alert was triggered because of: 'Saved Search [minor_host_not_reporting]: number of hosts(84)'
Recover padding is simply a line in the metadata file that gets created when Splunk needs to fix/recover metadata. Host values are an example of metadata, which are contained in a file called Hosts.data. You can simply ignore these names, although they can be a bit annoying when looking at your summary page.
To fix the summary page, you can modify the default summary dashboard to include the following additional query terms for the metadata search:
| search host!=*recover-padding*
So, any time you use a metadata search you will need to append the above, similar to this complete search: