Bugzilla

In bug 1693708 we cut out ~95 of the traffic (284k remain out of 5.3m events on 2021-02-19).

This can further be tailored by filtering out events that we believe to bear no correlation with YSODs.
Those may still be useful but we have now two days of full data to analyze if we want to, and we don't need to continue collecting so much.

In  this bug I'm going to cut out:
 - xhtml && NS_ERROR_PARSED_DATA_CACHED
 - xhtml && aboutNetError.xhtml
 - xhtml not coming from `omni.ja!`

I can see some of that data still being worth exploring, with questions such as:
 - why is aboutNetError.xhtml coming with so many errors of so many kinds at such a high volume?
 - what is NS_ERROR_PARSED_DATA_CACHED any why does it happen so often?
 - are the non-omni.ja XHTMLs cause YSODs that we filter out?

in particular, some of those events happen around YSOD. So when I see an YSOD and I query for zero_byte_load for that session_id I see multiple different zero_byte_loads *and* one on the file I saw YSOD on.
That may indicate that when I/O breaks it breaks on multiple files.

But for our YSOD search we don't need to continue collecting it. I recommend anyone looking for causes to always check a session from 2021-02-19 for all zero_byte_loads to see ones that we're now going to filter out.

In bug 1693708 we cut out ~95% of the traffic (284k remain out of 5.3m events on 2021-02-19).

This can further be tailored by filtering out events that we believe to bear no correlation with YSODs.
Those may still be useful but we have now two days of full data to analyze if we want to, and we don't need to continue collecting so much.

In  this bug I'm going to cut out:
 - xhtml && NS_ERROR_PARSED_DATA_CACHED
 - xhtml && aboutNetError.xhtml
 - xhtml not coming from `omni.ja!`

I can see some of that data still being worth exploring, with questions such as:
 - why is aboutNetError.xhtml coming with so many errors of so many kinds at such a high volume?
 - what is NS_ERROR_PARSED_DATA_CACHED any why does it happen so often?
 - are the non-omni.ja XHTMLs cause YSODs that we filter out?

in particular, some of those events happen around YSOD. So when I see an YSOD and I query for zero_byte_load for that session_id I see multiple different zero_byte_loads *and* one on the file I saw YSOD on.
That may indicate that when I/O breaks it breaks on multiple files.

But for our YSOD search we don't need to continue collecting it. I recommend anyone looking for causes to always check a session from 2021-02-19 for all zero_byte_loads to see ones that we're now going to filter out.