user/dev discussion of public-inbox itself
 help / color / mirror / code / Atom feed
From: Eric Wong <e@80x24.org>
To: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
Cc: meta@public-inbox.org
Subject: [PATCH 2/1] daemons: revamp periodic cleanup task
Date: Thu, 23 Sep 2021 00:46:25 +0000	[thread overview]
Message-ID: <20210923004625.GA23888@dcvr> (raw)
In-Reply-To: <20210922094517.GA8771@dcvr>

Neither Inboxes nor ExtSearch objects were retrying correctly
when there are live git processes, but the inboxes were getting
rescanned for search or other reasons.  Ensure the scan retries
eventually if there's live processes.

We also need to update the cleanup task to detect Xapian shard
count changes, since Xapian ->reopen is enough to detect any
other Xapian changes.  Otherwise, we just issue an inexpensive
->reopen call and let Xapian check whether there's anything
worth reopening.

This also lets us eliminate the Devel::Peek dependency.
---
 ci/deps.perl                 |  5 ---
 lib/PublicInbox/Admin.pm     |  3 +-
 lib/PublicInbox/ExtSearch.pm |  7 +++-
 lib/PublicInbox/Git.pm       | 11 ++++--
 lib/PublicInbox/Inbox.pm     | 67 ++++++++++++------------------------
 lib/PublicInbox/Search.pm    | 16 ++++++++-
 t/extsearch.t                | 15 ++++++++
 7 files changed, 68 insertions(+), 56 deletions(-)

diff --git a/ci/deps.perl b/ci/deps.perl
index a797911a..ae85986d 100755
--- a/ci/deps.perl
+++ b/ci/deps.perl
@@ -17,7 +17,6 @@ my $profiles = {
 	essential => [ qw(
 		git
 		perl
-		Devel::Peek
 		Digest::SHA
 		Encode
 		ExtUtils::MakeMaker
@@ -79,10 +78,6 @@ my $non_auto = {
 		pkg => 'p5-TimeDate',
 		rpm => 'perl-TimeDate',
 	},
-	'Devel::Peek' => {
-		deb => 'perl', # libperl5.XX, but the XX varies
-		pkg => 'perl5',
-	},
 	'Digest::SHA' => {
 		deb => 'perl', # libperl5.XX, but the XX varies
 		pkg => 'perl5',
diff --git a/lib/PublicInbox/Admin.pm b/lib/PublicInbox/Admin.pm
index 20964f9c..dcf17cf5 100644
--- a/lib/PublicInbox/Admin.pm
+++ b/lib/PublicInbox/Admin.pm
@@ -198,8 +198,7 @@ sub resolve_inboxes ($;$$) {
 	$opt->{-eidx_ok} ? (\@ibxs, \@eidx) : @ibxs;
 }
 
-# TODO: make Devel::Peek optional, only used for daemon
-my @base_mod = qw(Devel::Peek);
+my @base_mod = ();
 my @over_mod = qw(DBD::SQLite DBI);
 my %mod_groups = (
 	-index => [ @base_mod, @over_mod ],
diff --git a/lib/PublicInbox/ExtSearch.pm b/lib/PublicInbox/ExtSearch.pm
index cd7cba2e..7520e71c 100644
--- a/lib/PublicInbox/ExtSearch.pm
+++ b/lib/PublicInbox/ExtSearch.pm
@@ -112,6 +112,11 @@ sub description {
 		'$EXTINDEX_DIR/description missing';
 }
 
+sub search {
+	PublicInbox::Inbox::_cleanup_later($_[0]);
+	$_[0];
+}
+
 no warnings 'once';
 *base_url = \&PublicInbox::Inbox::base_url;
 *smsg_eml = \&PublicInbox::Inbox::smsg_eml;
@@ -121,6 +126,6 @@ no warnings 'once';
 *recent = \&PublicInbox::Inbox::recent;
 
 *max_git_epoch = *nntp_usable = *msg_by_path = \&mm; # undef
-*isrch = *search = \&PublicInbox::Search::reopen;
+*isrch = \&search;
 
 1;
diff --git a/lib/PublicInbox/Git.pm b/lib/PublicInbox/Git.pm
index cf51239f..3c577ab3 100644
--- a/lib/PublicInbox/Git.pm
+++ b/lib/PublicInbox/Git.pm
@@ -400,7 +400,7 @@ sub cleanup {
 	delete $self->{inflight_c};
 	_destroy($self, qw(cat_rbuf in out pid));
 	_destroy($self, qw(chk_rbuf in_c out_c pid_c err_c));
-	!!($self->{pid} || $self->{pid_c});
+	defined($self->{pid}) || defined($self->{pid_c});
 }
 
 
@@ -523,18 +523,25 @@ sub manifest_entry {
 	$ent;
 }
 
+# returns true if there are pending cat-file processes
 sub cleanup_if_unlinked {
 	my ($self) = @_;
 	return cleanup($self) if $^O ne 'linux';
 	# Linux-specific /proc/$PID/maps access
 	# TODO: support this inside git.git
+	my $ret = 0;
 	for my $fld (qw(pid pid_c)) {
 		my $pid = $self->{$fld} // next;
-		open my $fh, '<', "/proc/$pid/maps" or next;
+		open my $fh, '<', "/proc/$pid/maps" or return cleanup($self);
 		while (<$fh>) {
+			# n.b. we do not restart for unlinked multi-pack-index
+			# since it's not too huge, and the startup cost may
+			# be higher.
 			return cleanup($self) if /\.(?:idx|pack) \(deleted\)$/;
 		}
+		++$ret;
 	}
+	$ret;
 }
 
 1;
diff --git a/lib/PublicInbox/Inbox.pm b/lib/PublicInbox/Inbox.pm
index 02ffc7be..c0962af9 100644
--- a/lib/PublicInbox/Inbox.pm
+++ b/lib/PublicInbox/Inbox.pm
@@ -16,70 +16,47 @@ use Carp qw(croak);
 # admin will attempt to replace them atomically after compact/vacuum
 # and we need to be prepared for that.
 my $cleanup_timer;
-my $cleanup_avail = -1; # 0, or 1
-my $have_devel_peek;
 my $CLEANUP = {}; # string(inbox) -> inbox
 
 sub git_cleanup ($) {
 	my ($self) = @_;
-	if ($self->isa(__PACKAGE__)) {
-		# normal Inbox; low startup cost, and likely to have many
-		# many so this keeps process/pipe counts in check
-		$self->{git}->cleanup if $self->{git};
-	} else {
-		# ExtSearch, high startup cost if ->ALL, and probably
-		# only one per-daemon, so teardown only if required:
-		$self->git->cleanup_if_unlinked;
-	}
+	my $git = $self->{git} // return undef;
+	# normal inboxes have low startup cost and there may be many, so
+	# keep process+pipe counts in check.  ExtSearch may have high startup
+	# cost (e.g. ->ALL) and but likely one per-daemon, so cleanup only
+	# if there's unlinked files
+	my $live = $self->isa(__PACKAGE__) ? $git->cleanup
+					: $git->cleanup_if_unlinked;
+	delete($self->{git}) unless $live;
+	$live;
 }
 
+# returns true if further checking is required
+sub cleanup_shards { $_[0]->{search} ? $_[0]->{search}->cleanup_shards : undef }
+
 sub cleanup_task () {
 	$cleanup_timer = undef;
 	my $next = {};
 	for my $ibx (values %$CLEANUP) {
-		my $again;
-		if ($have_devel_peek) {
-			foreach my $f (qw(search)) {
-				# we bump refcnt by assigning tmp, here:
-				my $tmp = $ibx->{$f} or next;
-				next if Devel::Peek::SvREFCNT($tmp) > 2;
-				delete $ibx->{$f};
-				# refcnt is zero when tmp is out-of-scope
-			}
-		}
-		git_cleanup($ibx);
-		if (my $gits = $ibx->{-repo_objs}) {
-			foreach my $git (@$gits) {
-				$again = 1 if $git->cleanup;
-			}
+		my $again = git_cleanup($ibx);
+		$ibx->cleanup_shards and $again = 1;
+		for my $git (@{$ibx->{-repo_objs}}) {
+			$again = 1 if $git->cleanup;
 		}
 		check_inodes($ibx);
-		if ($have_devel_peek) {
-			$again ||= !!$ibx->{search};
-		}
 		$next->{"$ibx"} = $ibx if $again;
 	}
 	$CLEANUP = $next;
+	$cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
 }
 
-sub cleanup_possible () {
+sub _cleanup_later ($) {
 	# no need to require DS, here, if it were enabled another
 	# module would've require'd it, already
-	eval { PublicInbox::DS::in_loop() } or return 0;
-
-	eval {
-		require Devel::Peek; # needs separate package in Fedora
-		$have_devel_peek = 1;
-	};
-	1;
-}
-
-sub _cleanup_later ($) {
-	my ($self) = @_;
-	$cleanup_avail = cleanup_possible() if $cleanup_avail < 0;
-	return if $cleanup_avail != 1;
-	$cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
-	$CLEANUP->{"$self"} = $self;
+	if (eval { PublicInbox::DS::in_loop() }) {
+		$cleanup_timer //= PublicInbox::DS::later(\&cleanup_task);
+		$CLEANUP->{"$_[0]"} = $_[0]; # $self
+	}
 }
 
 sub _set_limiter ($$$) {
diff --git a/lib/PublicInbox/Search.pm b/lib/PublicInbox/Search.pm
index 1d1cd2f5..d285c11c 100644
--- a/lib/PublicInbox/Search.pm
+++ b/lib/PublicInbox/Search.pm
@@ -199,7 +199,7 @@ sub xdb_shards_flat ($) {
 	my (@xdb, $slow_phrase);
 	load_xapian();
 	$self->{qp_flags} //= $QP_FLAGS;
-	if ($xpfx =~ m/xapian${\SCHEMA_VERSION}\z/) {
+	if ($xpfx =~ m!/xapian[0-9]+\z!) {
 		@xdb = ($X{Database}->new($xpfx));
 		$self->{qp_flags} |= FLAG_PHRASE() if !-f "$xpfx/iamchert";
 	} else {
@@ -243,6 +243,20 @@ sub xdb ($) {
 	};
 }
 
+# returns true if a future rescan is desired
+sub cleanup_shards {
+	my ($self) = @_;
+	return unless exists($self->{xdb});
+	my $xpfx = $self->{xpfx};
+	return reopen($self) if $xpfx =~ m!/xapian[0-9]+\z!; # true
+	opendir(my $dh, $xpfx) or return warn("$xpfx gone: $!\n"); # true
+	my $nr = grep(/\A[0-9]+\z/, readdir($dh)) or
+		return warn("$xpfx has no shards\n"); # true
+	return reopen($self) if $nr == ($self->{nshard} // -1);
+	delete($self->{xdb});
+	undef;
+}
+
 sub new {
 	my ($class, $ibx) = @_;
 	ref $ibx or die "BUG: expected PublicInbox::Inbox object: $ibx";
diff --git a/t/extsearch.t b/t/extsearch.t
index ad4f2c6d..6c074022 100644
--- a/t/extsearch.t
+++ b/t/extsearch.t
@@ -436,12 +436,27 @@ for my $j (1, 3, 6) {
 
 SKIP: {
 	my $d = "$home/extindex-j1";
+	my $es = PublicInbox::ExtSearch->new($d);
+	ok(my $nresult0 = $es->mset('z:0..')->size, 'got results');
+	ok(ref($es->{xdb}), '{xdb} created');
+	my $nshards1 = $es->{nshard};
+	is($nshards1, 1, 'correct shard count');
+	my $xdb_str = "$es->{xdb}";
+	ok($es->cleanup_shards, 'cleanup_shards noop');
+	is("$es->{xdb}", $xdb_str, '{xdb} unchanged');
+
 	my $o = { 2 => \(my $err = '') };
 	ok(run_script([qw(-xcpdb -R4), $d]), 'xcpdb R4');
 	my @dirs = glob("$d/ei*/?");
 	for my $i (0..3) {
 		is(grep(m!/ei[0-9]+/$i\z!, @dirs), 1, "shard [$i] created");
 	}
+	is($es->cleanup_shards, undef, 'cleanup_shards cleaned');
+	ok(!defined($es->{xdb}), 'old {xdb} gone');
+	is($es->cleanup_shards, undef, 'cleanup_shards clean idempotent');
+	is($es->mset('z:0..')->size, $nresult0, 'new shards, same results');
+	ok($es->cleanup_shards, 'cleanup_shards true after open');
+
 	for my $i (4..5) {
 		is(grep(m!/ei[0-9]+/$i\z!, @dirs), 0, "no shard [$i]");
 	}

      parent reply	other threads:[~2021-09-23  0:46 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-21 14:47 Holding on to deleted packfiles Konstantin Ryabitsev
2021-09-21 19:06 ` Eric Wong
2021-09-21 20:57   ` Konstantin Ryabitsev
2021-09-21 21:37     ` Eric Wong
2021-09-22  9:45       ` [PATCH] gcf2 + extsearch: check for unlinked files on Linux Eric Wong
2021-09-22 19:51         ` Eric Wong
2021-09-23  0:46         ` Eric Wong [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://public-inbox.org/README

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210923004625.GA23888@dcvr \
    --to=e@80x24.org \
    --cc=konstantin@linuxfoundation.org \
    --cc=meta@public-inbox.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this 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).