Malformed output when log contains an equals sign

[tigerchase96]

Describe the bug
Whenever a container's console output contains a = b, Dozzle seems to misinterpret the line as a series of variables and ends up malforming the entire line.

To Reproduce

1. Run some variation of the following code, in any container. This example is in Python but the language doesn't matter. The only important part is the second print statement.

import time

print("hello world")
print("Failed to run somecommand with somevariable = 1")

while True:
    time.sleep(1)

2. Open the Dozzle logs for the container
3. Observe that the output looks like the screenshot below

Expected behavior
I expected the message to be printed as-is: Failed to run somecommand with somevariable = 1

Screenshots

Desktop (please complete the following information):

• OS: Windows 10
• Docker version: 25.0.3
• Browser & version: Chrome 122.0.6261.129
• Version: v6.3.1

[amir20]

Hmm that does seem like a tricky bug. Dozzle supports https://grafana.com/docs/loki/latest/send-data/promtail/stages/logfmt/ but the parser seems pretty relaxed.

I am not sure how to fix it.

[amir20]

Looks like I do strings.Contains(message, "="). Maybe I should change it to some regex without spaces around =.

[ncrothers]

Would it be reasonable to check if the message matches regex something like ^[^ ]+=[^ ].*? This would expect the message to start with a key that contains no spaces (which I think is a reasonable restriction), which I would imagine is a typical use case for key-value logs. That way it would match:

time=2012-11-01T22:08:41+00:00

But none of these:

This is a log message: "key=value"
Testing a variable = 2

[tigerchase96]

Thanks for the quick reply! If you can't find a good way to improve the regex, another option could be adding a new environment variable that disables all parsing and prints all messages as-is. Just tossing out an idea. 🙂

[amir20]

adding a new environment variable that disables all parsing and prints all messages as-is. Just tossing out an idea

Not worth it. Makes testing and configuration harder.

Let me play around with regex.

[amir20]

This should be fixed with #2835. Can you try it with amir20/dozzle:pr-2835?

[tigerchase96]

That appears to have fixed my problem. Just to be sure, I tested out some different variants to see what the output looks like:

print("Failed to run somecommand with somevariable = 1")
print("Failed to run somecommand with somevariable=1")
print("somevariable=1")
print("somevariable = 1")
print("somevariable=1 another=2")
print("somevariable = 1 another = 2")

My original test case is now correct, and somevariable=1 another=2 is still being parsed as expected.
I'm wondering if somevariable=1 should have been picked up by your updated regex, however.

[amir20]

somevariable=1 does not work because I check for at least two valid key value pairs. So I think it's ok. The reason is that a single log with key sounds like a bug. Will merge now.