Getting "_groktimeout" error for a particular filter sometimes. Need help optimizing the filter

[saurabhc]

The particular line is a "multiline". Hence using "codec => multiline" to get the line and then applying grok filter to it.
But sometimes when the particular line is having large content the parsing fails with "_groktimeout" error. logstash log file show WARN message like below.
Timeout executing grok '<grok filter>' against field 'message' with value 'Value too large to output (3138 bytes)! First 255 chars are: .....
Also, I see the cpu usage is 100% and java is using most of it.

I went through previous posts and do understand as my filter is a complicated one and has many DATA & GREEDYDATA patterns to match hence it might be taking more time. Need help with making a more efficient filter that returns me the same key-value pairs.

Below is my logstash config file.

input {
     file {
       type => "mylog"
       tags => "log"
	   path => [ "/Logs/mylog.log" ]
	   codec => multiline {
			pattern => "^\s"
			what => "previous"
	   }
       start_position => "beginning"
       sincedb_path => "NUL" 	   
     }
}
filter {
	 grok {
			  match => { "message" => "\[%{TIMESTAMP_ISO8601:log_timestamp}\] \[%{WORD:log_level}] \[tdid: %{NUMBER:tdid}\] \[userId: %{DATA:userId}\] \[myid: %{DATA:myid}\] \[%{TIMESTAMP_ISO8601:diagnostics_timestamp}\]\[p id:%{NUMBER:dpid}\]%{GREEDYDATA}(?<event_type>Start Data processing details)%{GREEDYDATA}Report path: (?<report_path>[^,\r\n]*)%{GREEDYDATA}Datamodel name: (?<datamodel_name>[^,\r\n]*)%{GREEDYDATA}User name: (?<user_name>[^,\r\n]*)%{GREEDYDATA}DB Session Module: (?<db_session_module>[^,\r\n]*)%{GREEDYDATA}DB Session Action: (?<db_session_action>[^,\r\n]*)%{GREEDYDATA}Total Number Of Rows Processed:(?<total_number_of_rows_processed>[^,\r\n]*)%{GREEDYDATA}Sql Action Id:(?<sql_action_id>[^,\r\n]*)%{GREEDYDATA}Total Number of Datasets: %{NUMBER:total_number_of_datasets}%{GREEDYDATA}Total Number of Elements: %{NUMBER:total_number_of_elements}%{GREEDYDATA}Total Parse Time: %{NUMBER:total_parse_time_ms}%{GREEDYDATA}Total Used Memory at Start: (?<total_used_memory_at_start>[^,\r\n]*)%{GREEDYDATA}Total Used Memory after Parsing: (?<total_used_memory_after_parsing>[^,\r\n]*)%{GREEDYDATA}Total Used Memory after Processing: (?<total_used_memory_after_processing>[^,\r\n]*)%{GREEDYDATA}Total Datamodel Validation Time: (?<total_datamodel_validation_time>[^,\r\n]*)%{GREEDYDATA}Total Xplan Generation Time: (?<total_xplan_generation_time>[^,\r\n]*)%{GREEDYDATA}Total Data Generation Time: (?<total_data_generation_time>[^,\r\n]*)%{GREEDYDATA}Total XML Data Size Generated: (?<total_xml_data_size_generated>[^,\r\n]*)%{GREEDYDATA}Processing Exceptions: (?<processing_exceptions>[^,\r\n]*)" }
      }
}

I've used "%{GREEDYDATA}" at multiple places because there could be no lines or multiple lines in between the patterns I'm trying to filter in the output.

Below is one particular matching line for which I got "_groktimeout" error.

"[2020-08-13T00:18:03.129+00:00] [INFO] [tdid: 136904] [userId: user1] [myid: 005f5ioH8qH3z035Rnh8id0003kz0001RV,1:33684] [2020-08-13T00:18:03.129+00:00][p id:1429221718][sch info:JobId:367679[RequestId:493608]][[
    Start Data processing details..
    ======================================================================
    processor Id: p{1429221718}
    Report path: /User/Human Capital Management/Interfaces/HCM_CLOUD_PROCESS
    Datamodel name: _User_Human_Capital_Management_Interfaces_HCM_CLOUD_PROCESS
    User name: user1
    DB Session Module: HCM_CLOUD_PROCESS
    DB Session Action: p{1429221718}{JobId:367679}
    key1: value key1
    key2: value key2
    key3: value key3
    key4: value key4
    key5: value key5
    key6: value key6
    key7: value key7
    key8: value key8
    key9: value key9
    key10: value key10
    key11: value key11
    key12: value key12
   .
   .
   .
   key_N: value key_N
        Total Number Of Times Query Executed:2
        Total Number Of Rows Processed:0
        Number of elements:0
        Total Query Execution time:1891
    ------------------------------------------------------------------------
    Total Number of Datasets: 5
    Total Number of Elements: 0
    Total Parse Time: 2
    Total Datamodel Validation Time: .000 mseconds
    Total Xplan Generation Time: .000 mseconds
    Total Data Generation Time: 2.000 seconds
    Total XML Data Size Generated: 1.03 KB
    Processing Exceptions: 
    ======================================================================
    End Data processing details..


"

[Badger]

I see no reason to use a single grok pattern to do this. You can match against multiple patterns. I have not tested this, but something like

grok {
    break_on_match => false
    match => {
        "message => [
            "Report path: (?<report_path>[^,\r\n]*)",
            "Datamodel name: (?<datamodel_name>[^,\r\n]*)",
            "User name: (?<user_name>[^,\r\n]*)",
            ...
        ]
    }
}

These patterns have fixed text at the beginning, so they will cheap to evaluate.

[saurabhc]

Thank You; it works.
I have few other problems because of this fix. Will create separate topic for that.