user/dev discussion of public-inbox itself
 help / color / mirror / code / Atom feed
Search results ordered by [date|relevance]  view[summary|nested|Atom feed]
thread overview below | download mbox.gz: |
* [PATCH 1/3] listener: EPOLL_CTL_ADD errors are non fatal
  2020-01-09 11:14  6% [PATCH 0/3] improve error handling for rare cases Eric Wong
@ 2020-01-09 11:14  7% ` Eric Wong
  0 siblings, 0 replies; 2+ results
From: Eric Wong @ 2020-01-09 11:14 UTC (permalink / raw)
  To: meta

EPOLL_CTL_ADD may fail with transient ENOMEM or ENOSPC errors,
so don't tear down the process when that happens.
---
 lib/PublicInbox/Listener.pm | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/lib/PublicInbox/Listener.pm b/lib/PublicInbox/Listener.pm
index 928d9301..cdfd34a2 100644
--- a/lib/PublicInbox/Listener.pm
+++ b/lib/PublicInbox/Listener.pm
@@ -40,7 +40,8 @@ sub event_step {
 	# on high-traffic sites.
 	if (my $addr = accept(my $c, $sock)) {
 		IO::Handle::blocking($c, 0); # no accept4 :<
-		$self->{post_accept}->($c, $addr, $sock);
+		eval { $self->{post_accept}->($c, $addr, $sock) };
+		warn "E: $@\n" if $@;
 		$self->requeue;
 	} elsif ($! == EAGAIN || $! == ECONNABORTED || $! == EPERM) {
 		# EAGAIN is common and likely

^ permalink raw reply related	[relevance 7%]

* [PATCH 0/3] improve error handling for rare cases
@ 2020-01-09 11:14  6% Eric Wong
  2020-01-09 11:14  7% ` [PATCH 1/3] listener: EPOLL_CTL_ADD errors are non fatal Eric Wong
  0 siblings, 1 reply; 2+ results
From: Eric Wong @ 2020-01-09 11:14 UTC (permalink / raw)
  To: meta

I tend to configure high RLIMIT_NOFILE limits, and the default
/proc/sys/fs/epoll/max_user_watches is difficult to hit, so
I've never hit the errors these patches supposedly fix...

But, I've noticed there's a places were we could die() without
an eval{} to trap it.  These patches fix some of those.

On a side note, we still lack sufficient testing for rare error
conditions.  These errors are not easy to reproduce, especially
EPOLL_CTL_ADD failures.

Also, the IO::KQueue->EV_SET calls actually make calls to
kqueue(2) directly (working more like epoll_ctl), so those could
die() inside the emulated epoll_ctl... Another day

Eric Wong (3):
  listener: EPOLL_CTL_ADD errors are non fatal
  http: log response_write errors
  qspawn: catch transient errors on pipe, EPOLL_CTL_ADD

 lib/PublicInbox/HTTP.pm     |  5 ++++-
 lib/PublicInbox/Listener.pm |  3 ++-
 lib/PublicInbox/Qspawn.pm   | 29 ++++++++++++++++++-----------
 3 files changed, 24 insertions(+), 13 deletions(-)

^ permalink raw reply	[relevance 6%]

Results 1-2 of 2 | reverse | options above
-- pct% links below jump to the message on this page, permalinks otherwise --
2020-01-09 11:14  6% [PATCH 0/3] improve error handling for rare cases Eric Wong
2020-01-09 11:14  7% ` [PATCH 1/3] listener: EPOLL_CTL_ADD errors are non fatal Eric Wong

Code repositories for project(s) associated with this public inbox

	https://80x24.org/public-inbox.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).