Stats sum command experiencing strange behavior after 7.2.0 upgrade

[paimonsoror]

Hi Folks;

So getting a very bizaare issue here after our upgrade to 7.2

index="app_rocket_dxs" sourcetype="fluentd_json" source="vbs-dxs-int*" 
| where message like "%Summary%" 
| eval temp=split(substr(message,64,250),":")  
| eval DomainName=mvindex(temp,1)  
| eval StartTime=mvindex(temp,3) 
| eval EndTime=mvindex(temp,5) 
| eval TopicName=mvindex(temp,7) 
| eval MsgCount=mvindex(temp,9) 
| convert num(MsgCount) as MsgCounts |convert timeformat="%Y-%m-%d" ctime(_time) AS date
| table StartTime,EndTime,MsgCounts,DomainName,TopicName,date
| stats sum(MsgCounts) as PublishedCount by date,TopicName
| sort date desc

Here is the way the data looks as a table

However after applying the stats command the 'PublishedCount' is blank:

Sample Event (scrubbed)

{"docker":{"container_id":"8203837773d4f65d9a3382d381c97f64af01209f865463239e7d59e6ed2972ec"},"kubernetes":{"container_name":"coverageemclntbenplan","namespace_name":"vbs-dxs-int","pod_name":"covplan-1-m9g6x","pod_id":"a8004109-e37d-11e8-b28e-fa163e193d33","labels":{"app":"covenplan","appname":"Rocket","deployment":"covernplan-1","deploymentconfig":"coveeplan"},"host":"cilver.com","master_url":"https://kubernetes.default.svc.cluster.local","namespace_id":"23eecb03-7947-11e8-9035-fa163ee5bb62"},"message":"11-12 16:37 oraclepool.oraclekafka INFO     Publisher Summary - Domain:coverage:Start_Bound:2018-11-12-11.33.26.421532 :End_Bound:2018-11-12-11.35.26.532198 :Topic Name:ibis.cif_em_o.em_clnt_ben_plan:count:0\n","level":"info","pipeline_metadata":{"collector":{"ipaddr4":"100.00.00.00","ipaddr6":"fe80::0000:0000:0000:a728","inputname":"fluent-plugin-systemd","name":"fluentd","received_at":"2018-11-12T16:37:21.820821+00:00","version":"0.12.43 1.6.0"}},"@timestamp":"2018-11-12T16:37:21.767889+00:00","viaq_index_name":"project.vbs-dxs-int.23eecb03-7947-11e8-9035-fa163ee5bb62.2018.11.12","viaq_msg_id":"NzM0OWEzZGEtMmJiNy00MDQ3LWI4ZjAtZTdkMGU1MzY0MzZj","kubernetes_node":"cilver.com"}

[martin_mueller · Accepted solution]

Look at your message string in detail - it has a tailing line breaker \n. Your approach of splitting at colons will include the line breaker in your field value, breaking the sum. Sanitize your value using trim() or replace(), or use rex instead of splitting:

... | rex field=message ":count:(?<MsgCounts>\d+)"

Side note, add the word Summary to your initial search to reduce the number of events loaded off disk (scanCount in the job inspector).

[paimonsoror]

you sir, are a scholar. thanks!

[martin_mueller]

Do post a sample event.

[paimonsoror]

Updated with sample!