[
https://issues.apache.org/jira/browse/TS-3213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14226291#comment-14226291
]
Sudheer Vinukonda commented on TS-3213:
---------------------------------------
Hi [~zwoop] - Apologies, I realize that my description of the issue is not very
clear. I didn't mean to say log buffer size is completely useless :=) - only
that it is useless in the specific scenario where the body factory is being
resolved for log fields. For the normal cases (e.g. squid/custom/access logging
for non-error scenarios), the configured log buffer sizes work well and as
expected.
{code}
[Nov 25 23:19:30.343] Server {0x2aeb9b5aa700} NOTE: Traffic Server is skipping
the current log entry because its size exceeds the maximum line (entry) size
for an ascii log buffer
{code}
So, the issue that we are mainly concerned right now is with a flood of the
below warning messages. These warnings are thrown when there are custom log
fields in the body factory templates (For e.g. we have a modified template for
{{urlrouting#no_mapping}} with a bunch of custom log fields along with
{{%<cquuc>}} which resolves to very long URLs much longer than the hard-coded
8K limit). And unfortunately enough, we have a ton of 404s generated (which in
itself, perhaps, needs to be investigated, if they are legit). I will discuss
more with you on how best to solve this problem.
> hardcoded log buffer size resulting in flood of warnings in diags.log despite
> configuring large log buffer size
> ---------------------------------------------------------------------------------------------------------------
>
> Key: TS-3213
> URL: https://issues.apache.org/jira/browse/TS-3213
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 5.1.1
> Reporter: Sudheer Vinukonda
> Fix For: 5.3.0
>
>
> We have seen a flood of the below warnings in diags.log, despite configuring
> the log buffer sizes sufficiently large. This happens even when we turned off
> ascii logging. Upon further investigation, it looks like, the body factory
> code tries to resolve the body factory templates as custom log entries
> (since, they may embed custom log fields), but, there's a hard-coded size of
> 8K for the buffer size that does the log field resolution at
> {{https://github.com/apache/trafficserver/blob/master/proxy/logging/LogAccess.cc#L1449}}.
> This makes the configured buffer values useless and results in throwing up
> the flood of warnings.
> {code}
> -bash-4.1$ sudo traffic_line -m . | grep log | grep size
> proxy.config.log.log_buffer_size 26624
> proxy.config.log.max_line_size 106496
> proxy.config.log.ascii_buffer_size 106496
> {code}
> {code}
> [Nov 25 23:19:30.343] Server {0x2aeb9b5aa700} NOTE: Traffic Server is
> skipping the current log entry because its size exceeds the maximum line
> (entry) size for an ascii log buffer
> {code}
> A simple fix is to use one of the configured log buffer sizes (preferrably,
> {{proxy.config.log.ascii_buffer_size}}) instead of the hard-coded 8k.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
