Google chat notifications are not received when bulk

[diegocejasprieto]

When firing multiples alerts at the same time for the same endpoint, only the first one is received in google chat. The rest of them seems to be discared
The pods shows errors like "error sending message","error":"non ok response from gchat" which does not tell you what's exactly going on.

The app should include a mechanism for retrying sending alerts when this happens or fire them after a few seconds between each other to avoid these kind of errors

I'm using image ghcr.io/mr-karan/calert:v2.1.1, deployed by helm chart

{"time":"2024-11-08T15:23:33.800146915Z","level":"INFO","source":{"function":"github.com/mr-karan/calert/internal/notifier.(*Notifier).Dispatch","file":"/home/runner/work/calert/calert/internal/notifier/notifier.go","line":41},"msg":"dispatching alerts","count":2}
{"time":"2024-11-08T15:23:33.800173642Z","level":"INFO","source":{"function":"github.com/mr-karan/calert/internal/providers/google_chat.(*GoogleChatManager).Push","file":"/home/runner/work/calert/calert/internal/providers/google_chat/google_chat.go","line":109},"msg":"dispatching alerts to google chat","count":2}
{"time":"2024-11-08T15:23:33.800273706Z","level":"INFO","source":{"function":"github.com/mr-karan/calert/internal/notifier.(*Notifier).Dispatch","file":"/home/runner/work/calert/calert/internal/notifier/notifier.go","line":41},"msg":"dispatching alerts","count":1}
{"time":"2024-11-08T15:23:33.800346855Z","level":"INFO","source":{"function":"github.com/mr-karan/calert/internal/providers/google_chat.(*GoogleChatManager).Push","file":"/home/runner/work/calert/calert/internal/providers/google_chat/google_chat.go","line":109},"msg":"dispatching alerts to google chat","count":1}
{"time":"2024-11-08T15:23:33.877335705Z","level":"ERROR","source":{"function":"github.com/mr-karan/calert/internal/providers/google_chat.(*GoogleChatManager).Push","file":"/home/runner/work/calert/calert/internal/providers/google_chat/google_chat.go","line":140},"msg":"error sending message","error":"non ok response from gchat"}
{"time":"2024-11-08T15:23:33.957092268Z","level":"ERROR","source":{"function":"github.com/mr-karan/calert/internal/providers/google_chat.(*GoogleChatManager).Push","file":"/home/runner/work/calert/calert/internal/providers/google_chat/google_chat.go","line":140},"msg":"error sending message","error":"non ok response from gchat"

[diegocejasprieto]

Additional info: I've put the pod in debug mode and got the following:

{"time":"2024-11-11T12:46:37.226655855Z","level":"DEBUG","source":{"function":"github.com/mr-karan/calert/internal/providers/google_chat.(*GoogleChatManager).sendMessage","file":"/home/runner/work/calert/calert/internal/providers/google_chat/message.go","line":110},"msg":"Non OK HTTP Response received from Google Chat Webhook endpoint","status":429,"responseBody":"{\n  \"error\": {\n    \"code\": 429,\n    \"message\": \"Resource has been exhausted (e.g. check quota).\",\n    \"status\": \"RESOURCE_EXHAUSTED\"\n  }\n}\n"}

I did some research and it seems that the Google API cannot handle multiple messages fired up at the same time (I didn't hit any other "quota" like amount of messages per hour or something).
That's why Calert should have a mechanism to retry those alerts instead of being discarded

[e100]

I have confirmed that it is related to how quickly messages are sent to the Google API.
I tried a delay of 50ms between each message, still had the problem so I tried 500ms and that seems to work.

diff --git a/internal/providers/google_chat/google_chat.go b/internal/providers/google_chat/google_chat.go
index 953c84d..df07d37 100644
--- a/internal/providers/google_chat/google_chat.go
+++ b/internal/providers/google_chat/google_chat.go
@@ -143,6 +143,7 @@ func (m *GoogleChatManager) Push(alerts []alertmgrtmpl.Alert) error {
                        }
 
                        m.metrics.Duration(fmt.Sprintf(`alerts_dispatched_duration_seconds{provider="%s", room="%s"}`, m.ID(), m.Room()), now)
+            time.Sleep(500 * time.Millisecond)
                }
        }

Google recommends to retry a few times when this happens with a longer delay between each try:
https://developers.google.com/workspace/chat/limits#resolve_time-based_quota_errors

[diegocejasprieto]

Hello Eric, so is this something that can't be set in Helm? If it isn't, can you submit a PR with your fix? Thx. El vie, 22 nov 2024 a la(s) 1:34 p.m., Eric Blevins ( ***@***.***) escribió:

…

I have confirmed that it is related to how quickly messages are sent to the Google API. I tried a delay of 50ms between each message, still had the problem so I tried 500ms and that seems to work. diff --git a/internal/providers/google_chat/google_chat.go b/internal/providers/google_chat/google_chat.go index 953c84d..df07d37 100644 --- a/internal/providers/google_chat/google_chat.go +++ b/internal/providers/google_chat/google_chat.go @@ -143,6 +143,7 @@ func (m *GoogleChatManager) Push(alerts []alertmgrtmpl.Alert) error { } m.metrics.Duration(fmt.Sprintf(`alerts_dispatched_duration_seconds{provider="%s", room="%s"}`, m.ID(), m.Room()), now) + time.Sleep(500 * time.Millisecond) } } Google recommends to retry a few times when this happens with a longer delay between each try: https://developers.google.com/workspace/chat/limits#resolve_time-based_quota_errors — Reply to this email directly, view it on GitHub <#83 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BIHURDAVQVBXODWCOVKG2JT2B5MH7AVCNFSM6AAAAABRNXH3OKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJUGE3DMOBWGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[allexf]

Same problem. And adding
time.Sleep(500 * time.Millisecond)
as shown in the post above, didn't help. Apparently a repeat mechanism is needed

[allexf]

I changed the sendMessage procedure in internal/providers/google_chat/message.go a bit to resend if an error is received. It doesn't look pretty, but it works for me.

func (m *GoogleChatManager) sendMessage(msg ChatMessage, threadKey string) error {
	out, err := json.Marshal(msg)
	if err != nil {
		return err
	}

	// Parse the webhook URL to add `?threadKey` param.
	u, err := url.Parse(m.endpoint)
	if err != nil {
		return err
	}
	q := u.Query()
	// Default behaviour is to start a new thread for every alert.
	q.Set("messageReplyOption", "MESSAGE_REPLY_OPTION_UNSPECIFIED")
	if m.threadedReplies {
		// If threaded replies are enabled, use the threadKey to reply to the same thread.
		q.Set("messageReplyOption", "REPLY_MESSAGE_FALLBACK_TO_NEW_THREAD")
		q.Set("threadKey", threadKey)
	}
	u.RawQuery = q.Encode()
	endpoint := u.String()

	i := 1
	for next := true; next; next = i < 5 {

	    // Prepare the request.
	    req, err := http.NewRequest("POST", endpoint, bytes.NewBuffer(out))
	    if err != nil {
		return err
	    }
	    req.Header.Set("Content-Type", "application/json")

	    // Send the request.
	    m.lo.Debug("sending alert", "url", endpoint, "msg", msg.Text)
	    resp, err := m.client.Do(req)
	    if err != nil {
		return err
	    }
	    defer resp.Body.Close()
	
	    if resp.StatusCode != http.StatusOK {
		time.Sleep(500 * time.Millisecond)
		fmt.Println("WARNING! HTTP Response Status:", resp.StatusCode, http.StatusText(resp.StatusCode),". Step:", i)
		i++
		if i == 4 {
		    // If response is non 200, log and throw the error.
		    // Read the response body
		    bodyBytes, err := io.ReadAll(resp.Body)
		    if err != nil {
			// Log the error if unable to read the response body
			m.lo.Error("Failed to read response body", "error", err)
			return fmt.Errorf("failed to read response body")
		    }
		    // Ensure the original response body is closed
		    defer resp.Body.Close()

		    // Convert the body bytes to a string for logging
		    responseBody := string(bodyBytes)

		    // Log the status code and response body at the debug level
		    m.lo.Debug("Non OK HTTP Response received from Google Chat Webhook endpoint", "status", resp.StatusCode, "responseBody", responseBody)

		    // Since the body has been read, if you need to use it later,
		    // you may need to reassign resp.Body with a new reader
		    resp.Body = io.NopCloser(bytes.NewBuffer(bodyBytes))

		    return fmt.Errorf("non ok response from gchat")
		}

	    } else {
		break
	    }
	}

	return nil
}

[PowerSurj]

@allexf would you be able to raise a PR for this, pease? this issue is bugging us.
TYVM!