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 01/11] watch: limit batch size of NNTP and IMAP workers, too
  2020-08-31  4:41  5% [PATCH 00/11] watch: fix contention w/ Maildir & NNTP Eric Wong
@ 2020-08-31  4:41  7% ` Eric Wong
  0 siblings, 0 replies; 2+ results
From: Eric Wong @ 2020-08-31  4:41 UTC (permalink / raw)
  To: meta; +Cc: Eric Wong

From: Eric Wong <e@yhbt.net>

We don't want to monopolize locks because processes can easily
block each other if using `watchspam' on a Maildir while a big
NNTP or IMAP import is happening.

This can also happen if somebody configured a single inbox to
watch from several sources to merge several mailboxes into one
(e.g. both an IMAP and Maildir are watched).
---
 lib/PublicInbox/WatchMaildir.pm | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/lib/PublicInbox/WatchMaildir.pm b/lib/PublicInbox/WatchMaildir.pm
index a227a6fd..5176ef69 100644
--- a/lib/PublicInbox/WatchMaildir.pm
+++ b/lib/PublicInbox/WatchMaildir.pm
@@ -108,6 +108,7 @@ sub new {
 	return unless $mdre || scalar(keys %imap) || scalar(keys %nntp);
 
 	bless {
+		max_batch => 10, # avoid hogging locks for too long
 		spamcheck => $spamcheck,
 		mdmap => \%mdmap,
 		mdre => $mdre,
@@ -472,8 +473,14 @@ sub imap_fetch_all ($$$) {
 
 		$l_uid = $uids->[-1] + 1; # for next search
 		my $last_uid;
+		my $n = $self->{max_batch};
 
 		while (scalar @$uids) {
+			if (--$n < 0) {
+				_done_for_now($self);
+				$itrk->update_last($r_uidval, $last_uid);
+				$n = $self->{max_batch};
+			}
 			my @batch = splice(@$uids, 0, $bs);
 			$batch = join(',', @batch);
 			local $0 = "UID:$batch $mbx $sec";
@@ -888,9 +895,15 @@ sub nntp_fetch_all ($$$) {
 	};
 	my $inboxes = $self->{nntp}->{$url};
 	my $last_art;
+	my $n = $self->{max_batch};
 	for ($beg..$end) {
 		last if $self->{quit};
 		$art = $_;
+		if (--$n < 0) {
+			_done_for_now($self);
+			$itrk->update_last(0, $last_art);
+			$n = $self->{max_batch};
+		}
 		my $raw = $nn->article($art);
 		unless (defined($raw)) {
 			my $msg = $nn->message;
@@ -976,12 +989,11 @@ sub fs_scan_step {
 	local $PublicInbox::DS::in_loop = 0; # waitpid() synchronously
 
 	# continue existing scan
-	my $max = 10;
 	my $opendirs = $self->{opendirs};
 	my @dirnames = keys %$opendirs;
 	foreach my $dir (@dirnames) {
 		my $dh = delete $opendirs->{$dir};
-		my $n = $max;
+		my $n = $self->{max_batch};
 		while (my $fn = readdir($dh)) {
 			_try_path($self, "$dir/$fn");
 			last if --$n < 0;
@@ -996,7 +1008,7 @@ sub fs_scan_step {
 				warn "failed to open $dir: $!\n";
 				next;
 			}
-			my $n = $max;
+			my $n = $self->{max_batch};
 			while (my $fn = readdir($dh)) {
 				_try_path($self, "$dir/$fn");
 				last if --$n < 0;

^ permalink raw reply related	[relevance 7%]

* [PATCH 00/11] watch: fix contention w/ Maildir & NNTP
@ 2020-08-31  4:41  5% Eric Wong
  2020-08-31  4:41  7% ` [PATCH 01/11] watch: limit batch size of NNTP and IMAP workers, too Eric Wong
  0 siblings, 1 reply; 2+ results
From: Eric Wong @ 2020-08-31  4:41 UTC (permalink / raw)
  To: meta

Here's a bunch of fixes to improve watch performance when
both Maildirs and NNTP are being watched (possibly on the same
inbox, or if `watchspam' is configured for spam removals).

Wakeups are reduced, and inbox.lock contention is minimized by
using read-only ->over to check for `watchspam' removals.

These affect IMAP, too; but I've been mainly using NNTP.

Eric Wong (11):
  watch: limit batch size of NNTP and IMAP workers, too
  watchmaildir: use v5.10.1, drop warnings
  rename WatchMaildir => Watch
  watch: log signal activities to STDERR
  watch: avoid unnecessary spawning on spam removals
  watch: block signals before fork on non-signalfd/kevent systems
  watch: comments and tiny cleanups
  ds: avoid excessive queueing when reaping PIDs
  watch: use EOFpipe to reduce dwaitpid wakeups
  ds: avoid unnecessary timer for waitpid
  replace ParentPipe with EOFpipe

 MANIFEST                                      |   4 +-
 lib/PublicInbox/DS.pm                         |  38 +++---
 lib/PublicInbox/Daemon.pm                     |   6 +-
 lib/PublicInbox/EOFpipe.pm                    |  24 ++++
 lib/PublicInbox/Import.pm                     |   3 +
 lib/PublicInbox/ParentPipe.pm                 |  23 ----
 lib/PublicInbox/V2Writable.pm                 |   3 +
 lib/PublicInbox/{WatchMaildir.pm => Watch.pm} | 111 +++++++++++++-----
 script/public-inbox-watch                     |  34 ++++--
 t/imapd.t                                     |   2 +-
 t/nntpd.t                                     |   2 +-
 t/watch_filter_rubylang.t                     |   4 +-
 t/watch_imap.t                                |   4 +-
 t/watch_maildir.t                             |  18 +--
 t/watch_maildir_v2.t                          |  22 ++--
 t/watch_multiple_headers.t                    |   4 +-
 t/watch_nntp.t                                |   4 +-
 17 files changed, 190 insertions(+), 116 deletions(-)
 create mode 100644 lib/PublicInbox/EOFpipe.pm
 delete mode 100644 lib/PublicInbox/ParentPipe.pm
 rename lib/PublicInbox/{WatchMaildir.pm => Watch.pm} (92%)

^ permalink raw reply	[relevance 5%]

Results 1-2 of 2 | reverse | options above
-- pct% links below jump to the message on this page, permalinks otherwise --
2020-08-31  4:41  5% [PATCH 00/11] watch: fix contention w/ Maildir & NNTP Eric Wong
2020-08-31  4:41  7% ` [PATCH 01/11] watch: limit batch size of NNTP and IMAP workers, too 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).