PureeBufferedOutput will never resume, if the flushSync function is skipped by locking storage with another PureeBufferedOutput instance

[sys1yagi]

Hi.

Problem

When there are two or more PureeBufferedOutputs, if the flushSync function is skipped by locking storage with another PureeBufferedOutput instance, it will never resume.

Caused by

• PureeBufferedOutput has RetryableTaskRunner and send log by the runner.
• RetryableTaskRunner start by tryToStart function but the function skip if already 'future' exists. ★
• PureeBufferedOutput call tryToStart function when receive new log. ★
• PureeBufferedOutput#flushSync() will be skipped when 'storage' is locked. ★
• RetryableTaskRunner resets 'future' only if it have done the send process (success / failed). ★
• So, if there are multiple PureeBufferedOutputs and one locks the 'storage', the other instance's "future" will not be reset and the flushTask will not work. 😵

Idea

I have no idea... 🙀

thanks.

[litmon]

@sys1yagi Thanks for reporting.
I patched this bug in #62 , but it cannot completely fix because if this situation loops it will be occurred.
In the future, I will try to split storage's lock which has each PureeBufferedOuput.