-
Notifications
You must be signed in to change notification settings - Fork 28
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Syslog drain app error messages in app log stream #633
base: main
Are you sure you want to change the base?
Syslog drain app error messages in app log stream #633
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@corporatemax Thanks for the PR and the time invested in it. Generally it looks good. I have only few minor comments and I would wait for the other approvers to tell what they think.
|
||
// WriteLog writes a message in the application log stream using a LogClient. | ||
func (appLogEmitter *LoggregatorEmitter) WriteLog(appID string, message string) { | ||
if appLogEmitter.logClient == nil || appID == "" { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wonder if we should write a warning in the Syslog Agent's logs that the emitter is not set and the app dev won't know that something went wrong... On the other hand, I understand that this is only to safeguard the EmitLog
call as everything is initialized properly in the syslog_agent.go cmd.
I feel like if we don't write a log message for the syslog agent, we are covering the misconfiguration... Of course, this might not be the write place as the log will be written on every WriteLog
call.
@ctlong Any thoughts on this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am not sure if the location of this file is sufficient as I am also calling it in the filtered binding fetcher. Should I move this in its separate package? If yes where exactly?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's leave it as is for now until we think of something better.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@chombium it sounds like you're worried that we do not have any mechanisms to prevent future misconfiguration of the AppLogEmitter
when it's initialized in syslog_agent.go
, is that right?
I think we could adequately cover this case with a new test in syslog_agent_test.go
. That would seem like a valuable addition to me.
"golang.org/x/net/context" | ||
"log" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Was this reformatting? If so please run golangci-lint run ./...
in the /src
folder
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
no this was goland - I ran the formatter on my side and commited the changes.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looks very strange to me. Usually, the standard libraries are grouped at the top... can you please double-check the formatting here, thanks
@@ -138,7 +122,7 @@ func (w *SyslogConnector) Connect(ctx context.Context, b Binding) (egress.Writer | |||
w.droppedMetric.Add(float64(missed)) | |||
drainDroppedMetric.Add(float64(missed)) | |||
|
|||
w.emitLoggregatorErrorLog(b.AppId, fmt.Sprintf("%d messages lost for application %s in user provided syslog drain with url %s", missed, b.AppId, anonymousUrl.String())) | |||
w.loggregatorEmitter.WriteLog(b.AppId, fmt.Sprintf("%d messages lost for application %s in user provided syslog drain with url %s", missed, b.AppId, anonymousUrl.String())) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a system/operator message and we should adjust the text as we have a concrete App for which some logs are lost. IMO x messages lost for syslog drain with url y
is enough. On the other hand if the same syslog drain url is used for multiple apps the message would be clearer this way. Though, the app id will be sent anyhow as label/tag of the log. I'm not totally sure if we should change the message.
@ctlong any thought on this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What is the final log message output to the user in this case? I think the log client, or some of the other logic in AppLogEmitter
may prefix the message with some information. In any case, you're right, we should evaluate the full log envelope and see if anything should be tweaked to ensure that it is legible to an app developer.
bf6dd19
to
9f789db
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@corporatemax Thanks for looking into the review and making adjustments. I'm happy with the how the things look like.
I wonder what @ctlong has to say about this PR.
@corporatemax Can you please check the unit tests? Do they fail because they run as GH Action or it's something else? |
@chombium the test states
I assume this test is flaky regarding the opened ports. |
@corporatemax I was able to reproduce the error locally on Ubuntu Linux 20.04 LTS when running both all the tests and only the failing test suite. All tests run:
and running the single suite:
At first I thought that we have some problem when all of the tests run together, but the problem is isolated, but it seems that the the problem is within the I've also tried running single test like:
and that run properly. It's seems to me there are problems with how the tests are setup and run when running the whole |
I've rerun the tests as they run in the GitHub action and I could reproduce the problem. Therefore I will revoke my approval until all test are running properly
} | ||
} | ||
|
||
func (f WriterFactory) NewWriter(ub *URLBinding) (egress.WriteCloser, error) { | ||
func (f WriterFactory) NewWriter(ub *URLBinding, emitter AppLogEmitter) (egress.WriteCloser, error) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What is the utility of providing WriterFactory
with an AppLogEmitter
and passing another AppLogEmitter
to WriterFactory.NewWriter()
? It seems to me like they are the same AppLogEmitter
and so could be de-duped, but maybe I'm missing something.
Description
This change should resolve #579 .
I pulled the SyslogConnector.emitLoggregatorErrorLog function into the app_log_emitter.go file and moved its dependent variables accordingly and capsuled it with the AppLogEmitter struct.
Furthermore, I used the AppLogEmitter in the RetryWriter and the WriterFactory to provide error messages to the Application Developer who tries to configure a syslog drain but does not know why there are no logs showing up in the syslog drain target.
I did this also in the FilteredBindingFetcher as there missconfigurations leads to errors there.
Type of change
Testing performed?
Checklist:
main
branch, or relevant version branchIf you have any questions, or want to get attention for a PR or issue please reach out on the #logging-and-metrics channel in the cloudfoundry slack