Closed jrgm closed 10 years ago
@whd would like it so that the default/backstop uid value be such that it can be distinguished between cases where not having a UID is ok (request not needing auth) vs. a request where auth is required but the uid could not be determined.
This is an example broken message:
Timestamp: 2014-07-09 23:46:52.256 +0000 UTC
Type: request.summary
Hostname: hostname
Pid: 1234
UUID: e179d273-4421-4fed-ab9b-f72e4243e4ad
Logger: FxaAuth
Payload:
EnvVersion: 0
Severity: 7
Fields: [name:"remoteAddressChain" value_string:"IP" value_string:"IP" value_string:"127.0.0.1" name:"path" value_string:"/v1/certificate/sign" name:"code" value_type:DOUBLE value_double:200 name:"errno" value_type:DOUBLE value_double:0 name:"msg" value_string:"" name:"user_agent_os" value_string:"Windows 8.1" name:"user_agent_version" value_type:DOUBLE value_double:30 name:"user_agent_browser" value_string:"Firefox" name:"name" value_string:"fxa-auth-server" name:"t" value_type:DOUBLE value_double:170 name:"rid" value_string:"1404949612086-2457-20264" name:"lang" value_string:"sv" name:"level" value_type:DOUBLE value_double:30 ]
"default" uid is '00'
uid value be such that it can be distinguished between cases where not having a UID is ok (request not needing auth) vs. a request where auth is required but the uid could not be determined.
The errno signals when a uid is required but not provided.
When we had that little incident on Jul 10 for auth-server, heka choked on log messages that had no uid field on path
/v1/certificate/sign
. We should make sure that the uid exists even in error cases (and set uid to some negative number?).@whd can you provide an example broken message?