We are using v9 format of logs in splunk. It is working fine and we are able to see logs in splunk as expected.
We added 4 more fields in transform.conf and test the addon in splunk. Then additional fields taking the value
of s3_filename, bucket name and prefix which are added at the end which is not correct behavior.
We are looking for solution with that we should be able to parse correct value in correct field and the additional fields should have null values if there is no values for them in logs.
transform.conf
[proxylogs_fields]
DELIMS = ","
FIELDS = Timestamp,policy_identities,src,src_translated_ip,dest,content_type,action,url,http_referrer,http_user_agent,status,requestSize,responseSize,responseBodySize,sha256,category,av_detection,pua,amp_disposition,amp_malwarename,amp_score,policy_identity_type,blocked_category,identities,identity_type,request_method,dlp_status,certificate_errors,filename,rulesetID,ruleID,destinationListID,isolateAction,fileAction,warnStatus,forwarding_method,Producer,test_feild1,test_field2,test_field3,test_field4,s3_filename,aws_bucket_name,aws_prefix
props.conf
[cisco:cloud_security:proxy]
REPORT-proxylogs-fields = proxylogs_fields,extract_url_domain
LINE_BREAKER = ([\r\n]+)
# EVENT_BREAKER = ([\r\n]+)
# EVENT_BREAKER_ENABLE = true
SHOULD_LINEMERGE = false
CHARSET = AUTO
disabled = false
TRUNCATE = 1000000
MAX_EVENTS = 1000000
EVAL-product = "Cisco Secure Access and Umbrella"
EVAL-vendor = "Cisco"
EVAL-vendor_product = "Cisco Secure Access/Umbrella"
MAX_TIMESTAMP_LOOKAHEAD = 22
NO_BINARY_CHECK = true
TIME_PREFIX = ^
TIME_FORMAT = "%Y-%m-%d %H:%M:%S"
TZ = UTC
FIELDALIAS-bytes_in = requestSize as bytes_in
FIELDALIAS-bytes_out = responseSize as bytes_out
EVAL-action = lower(action)
EVAL-app = "Cisco Cloud Security"
FIELDALIAS-http_content_type = content_type as http_content_type
EVAL-http_user_agent_length = len(http_user_agent)
EVAL-url_length = len(url)
EVAL-dest = if(isnotnull(dest),dest,url_domain)
EVAL-bytes = requestSize + responseSize
I think I understand what you are asking about but without sample ingested data and the new output sample it is harder to decipher what is going wrong.