From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on dcvr.yhbt.net X-Spam-Level: X-Spam-Status: No, score=-4.0 required=3.0 tests=ALL_TRUSTED,BAYES_00 shortcircuit=no autolearn=ham autolearn_force=no version=3.4.2 Received: from localhost (dcvr.yhbt.net [127.0.0.1]) by dcvr.yhbt.net (Postfix) with ESMTP id E7DF11F5AE; Fri, 31 Jul 2020 21:36:18 +0000 (UTC) Date: Fri, 31 Jul 2020 21:36:18 +0000 From: Eric Wong To: meta@public-inbox.org Subject: [PATCH 2/1 v2] improve error handling on import fork / lock failures Message-ID: <20200731213618.GA19572@dcvr> References: <20200730080533.GA8841@dcvr> <20200731085619.GA10081@dcvr> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <20200731085619.GA10081@dcvr> List-Id: v?fork failures seems to be the cause of locks not getting released in -watch. Ensure lock release doesn't get skipped in ->done for both v1 and v2 inboxes. We also need to do everything we can to ensure DB handles, pipes and processes get released even in the face of failure. While we're at it, make failures around `git update-server-info' non-fatal, since smart HTTP seems more popular anyways. v2 changes: - spawn: show failing command - ensure waitpid is synchronous for inotify events - teardown all fast-import processes on exception, not just the failing one - beef up lock_release error handling - release lock on fast-import spawn failure --- lib/PublicInbox/DirIdle.pm | 1 + lib/PublicInbox/Import.pm | 84 +++++++++++++++++++-------------- lib/PublicInbox/Lock.pm | 12 ++--- lib/PublicInbox/Spawn.pm | 2 +- lib/PublicInbox/V2Writable.pm | 28 ++++++++--- lib/PublicInbox/WatchMaildir.pm | 36 +++++++++----- t/psgi_search.t | 1 + 7 files changed, 101 insertions(+), 63 deletions(-) diff --git a/lib/PublicInbox/DirIdle.pm b/lib/PublicInbox/DirIdle.pm index 89cce305f..daa2212b4 100644 --- a/lib/PublicInbox/DirIdle.pm +++ b/lib/PublicInbox/DirIdle.pm @@ -44,6 +44,7 @@ sub new { sub event_step { my ($self) = @_; my $cb = $self->{cb}; + local $PublicInbox::DS::in_loop = 0; # waitpid() synchronously eval { my @events = $self->{inot}->read; # Linux::Inotify2->read $cb->($_) for @events; diff --git a/lib/PublicInbox/Import.pm b/lib/PublicInbox/Import.pm index b50c662c7..07a495187 100644 --- a/lib/PublicInbox/Import.pm +++ b/lib/PublicInbox/Import.pm @@ -48,32 +48,35 @@ sub gfi_start { return ($self->{in}, $self->{out}) if $self->{pid}; - my ($out_r, $out_w); + my (@ret, $out_r, $out_w); pipe($out_r, $out_w) or die "pipe failed: $!"; - my $git = $self->{git}; $self->lock_acquire; - - local $/ = "\n"; - my $ref = $self->{ref}; - chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref)); - if ($self->{path_type} ne '2/38' && $self->{tip}) { - local $/ = "\0"; - my @tree = $git->qx(qw(ls-tree -r -z --name-only), $ref); - chomp @tree; - $self->{-tree} = { map { $_ => 1 } @tree }; + eval { + my ($git, $ref) = @$self{qw(git ref)}; + local $/ = "\n"; + chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref)); + if ($self->{path_type} ne '2/38' && $self->{tip}) { + local $/ = "\0"; + my @t = $git->qx(qw(ls-tree -r -z --name-only), $ref); + chomp @t; + $self->{-tree} = { map { $_ => 1 } @t }; + } + my @cmd = ('git', "--git-dir=$git->{git_dir}", + qw(fast-import --quiet --done --date-format=raw)); + my ($in_r, $pid) = popen_rd(\@cmd, undef, { 0 => $out_r }); + $out_w->autoflush(1); + $self->{in} = $in_r; + $self->{out} = $out_w; + $self->{pid} = $pid; + $self->{nchg} = 0; + @ret = ($in_r, $out_w); + }; + if ($@) { + $self->lock_release; + die $@; } - - my $git_dir = $git->{git_dir}; - my @cmd = ('git', "--git-dir=$git_dir", qw(fast-import - --quiet --done --date-format=raw)); - my ($in_r, $pid) = popen_rd(\@cmd, undef, { 0 => $out_r }); - $out_w->autoflush(1); - $self->{in} = $in_r; - $self->{out} = $out_w; - $self->{pid} = $pid; - $self->{nchg} = 0; - ($in_r, $out_w); + @ret; } sub wfail () { die "write to fast-import failed: $!" } @@ -175,13 +178,16 @@ sub _update_git_info ($$) { my $env = { GIT_INDEX_FILE => $index }; run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env); } - run_die([@cmd, 'update-server-info']); + eval { run_die([@cmd, 'update-server-info']) }; my $ibx = $self->{ibx}; - ($ibx && $self->{path_type} eq '2/38') and eval { - require PublicInbox::SearchIdx; - my $s = PublicInbox::SearchIdx->new($ibx); - $s->index_sync({ ref => $self->{ref} }); - }; + if ($ibx && $ibx->version == 1 && -d "$ibx->{inboxdir}/public-inbox" && + eval { require PublicInbox::SearchIdx }) { + eval { + my $s = PublicInbox::SearchIdx->new($ibx); + $s->index_sync({ ref => $self->{ref} }); + }; + warn "$ibx->{inboxdir} index failed: $@\n" if $@; + } eval { run_die([@cmd, qw(gc --auto)]) } if $do_gc; } @@ -460,17 +466,23 @@ sub init_bare { sub done { my ($self) = @_; my $w = delete $self->{out} or return; - my $r = delete $self->{in} or die 'BUG: missing {in} when done'; - print $w "done\n" or wfail; - my $pid = delete $self->{pid} or die 'BUG: missing {pid} when done'; - waitpid($pid, 0) == $pid or die 'fast-import did not finish'; - $? == 0 or die "fast-import failed: $?"; - + eval { + my $r = delete $self->{in} or die 'BUG: missing {in} when done'; + print $w "done\n" or wfail; + my $pid = delete $self->{pid} or + die 'BUG: missing {pid} when done'; + waitpid($pid, 0) == $pid or die 'fast-import did not finish'; + $? == 0 or die "fast-import failed: $?"; + }; + my $wait_err = $@; my $nchg = delete $self->{nchg}; - _update_git_info($self, 1) if $nchg; + if ($nchg && !$wait_err) { + eval { _update_git_info($self, 1) }; + warn "E: $self->{git}->{git_dir} update info: $@\n" if $@; + } $self->lock_release(!!$nchg); - $self->{git}->cleanup; + die $wait_err if $wait_err; } sub atfork_child { diff --git a/lib/PublicInbox/Lock.pm b/lib/PublicInbox/Lock.pm index ca43682f8..b2c8227f0 100644 --- a/lib/PublicInbox/Lock.pm +++ b/lib/PublicInbox/Lock.pm @@ -16,20 +16,20 @@ sub lock_acquire { croak 'already locked '.($lock_path // '(undef)') if $self->{lockfh}; return unless defined($lock_path); sysopen(my $lockfh, $lock_path, O_WRONLY|O_CREAT) or - die "failed to open lock $lock_path: $!\n"; - flock($lockfh, LOCK_EX) or die "lock failed: $!\n"; + croak "failed to open $lock_path: $!\n"; + flock($lockfh, LOCK_EX) or croak "lock $lock_path failed: $!\n"; $self->{lockfh} = $lockfh; } sub lock_release { my ($self, $wake) = @_; - return unless $self->{lock_path}; - my $lockfh = delete $self->{lockfh} or croak 'not locked'; + defined(my $lock_path = $self->{lock_path}) or return; + my $lockfh = delete $self->{lockfh} or croak "not locked: $lock_path"; syswrite($lockfh, '.') if $wake; - flock($lockfh, LOCK_UN) or die "unlock failed: $!\n"; - close $lockfh or die "close failed: $!\n"; + flock($lockfh, LOCK_UN) or croak "unlock $lock_path failed: $!\n"; + close $lockfh or croak "close $lock_path failed: $!\n"; } 1; diff --git a/lib/PublicInbox/Spawn.pm b/lib/PublicInbox/Spawn.pm index 50f318515..508d43fd7 100644 --- a/lib/PublicInbox/Spawn.pm +++ b/lib/PublicInbox/Spawn.pm @@ -275,7 +275,7 @@ sub spawn ($;$$) { } my $cd = $opts->{'-C'} // ''; # undef => NULL mapping doesn't work? my $pid = pi_fork_exec($redir, $f, $cmd, \@env, $rlim, $cd); - die "fork_exec failed: $!\n" unless $pid > 0; + die "fork_exec @$cmd failed: $!\n" unless $pid > 0; $pid; } diff --git a/lib/PublicInbox/V2Writable.pm b/lib/PublicInbox/V2Writable.pm index e071bc1e0..e1c9a393a 100644 --- a/lib/PublicInbox/V2Writable.pm +++ b/lib/PublicInbox/V2Writable.pm @@ -660,21 +660,35 @@ sub barrier { checkpoint($_[0], 1) }; # public sub done { my ($self) = @_; - my $im = delete $self->{im}; - $im->done if $im; # PublicInbox::Import::done - checkpoint($self); - my $mm = delete $self->{mm}; - $mm->{dbh}->commit if $mm; + my $err = ''; + if (my $im = delete $self->{im}) { + eval { $im->done }; # PublicInbox::Import::done + $err .= "import done: $@\n" if $@; + } + if (!$err) { + eval { checkpoint($self) }; + $err .= "checkpoint: $@\n" if $@; + } + if (my $mm = delete $self->{mm}) { + my $m = $err ? 'rollback' : 'commit'; + eval { $mm->{dbh}->$m }; + $err .= "msgmap $m: $@\n" if $@; + } my $shards = delete $self->{idx_shards}; if ($shards) { - $_->remote_close for @$shards; + for (@$shards) { + eval { $_->remote_close }; + $err .= "shard close: $@\n" if $@; + } } - $self->{over}->disconnect; + eval { $self->{over}->disconnect }; + $err .= "over disconnect: $@\n" if $@; delete $self->{bnote}; my $nbytes = $self->{total_bytes}; $self->{total_bytes} = 0; $self->lock_release(!!$nbytes) if $shards; $self->{ibx}->git->cleanup; + die $err if $err; } sub fill_alternates ($$) { diff --git a/lib/PublicInbox/WatchMaildir.pm b/lib/PublicInbox/WatchMaildir.pm index 7547f6e47..fad708d8f 100644 --- a/lib/PublicInbox/WatchMaildir.pm +++ b/lib/PublicInbox/WatchMaildir.pm @@ -124,8 +124,10 @@ sub new { sub _done_for_now { my ($self) = @_; local $PublicInbox::DS::in_loop = 0; # waitpid() synchronously - for (values %{$self->{importers}}) { - $_->done if $_; # $_ may be undef during cleanup + for my $im (values %{$self->{importers}}) { + next if !$im; # $im may be undef during cleanup + eval { $im->done }; + warn "$im->{ibx}->{name} ->done: $@\n" if $@; } } @@ -137,12 +139,15 @@ sub remove_eml_i { # each_inbox callback $im->remove($eml, 'spam'); if (my $scrub = $ibx->filter($im)) { my $scrubbed = $scrub->scrub($eml, 1); - $scrubbed or return; - $scrubbed == REJECT() and return; - $im->remove($scrubbed, 'spam'); + if ($scrubbed && $scrubbed != REJECT) { + $im->remove($scrubbed, 'spam'); + } } }; - warn "error removing spam at: $loc from $ibx->{name}: $@\n" if $@; + if ($@) { + warn "error removing spam at: $loc from $ibx->{name}: $@\n"; + _done_for_now($self); + } } sub _remove_spam { @@ -155,7 +160,6 @@ sub _remove_spam { sub import_eml ($$$) { my ($self, $ibx, $eml) = @_; - my $im = _importer_for($self, $ibx); # any header match means it's eligible for the inbox: if (my $watch_hdrs = $ibx->{-watchheaders}) { @@ -167,13 +171,19 @@ sub import_eml ($$$) { } return unless $ok; } - - if (my $scrub = $ibx->filter($im)) { - my $ret = $scrub->scrub($eml) or return; - $ret == REJECT() and return; - $eml = $ret; + eval { + my $im = _importer_for($self, $ibx); + if (my $scrub = $ibx->filter($im)) { + my $scrubbed = $scrub->scrub($eml) or return; + $scrubbed == REJECT and return; + $eml = $scrubbed; + } + $im->add($eml, $self->{spamcheck}); + }; + if ($@) { + warn "$ibx->{name} add failed: $@\n"; + _done_for_now($self); } - $im->add($eml, $self->{spamcheck}); } sub _try_path { diff --git a/t/psgi_search.t b/t/psgi_search.t index 64f8b1ac5..2d12ba6a2 100644 --- a/t/psgi_search.t +++ b/t/psgi_search.t @@ -14,6 +14,7 @@ my @mods = qw(DBD::SQLite Search::Xapian HTTP::Request::Common Plack::Test require_mods(@mods); use_ok($_) for (qw(HTTP::Request::Common Plack::Test)); use_ok 'PublicInbox::WWW'; +use_ok 'PublicInbox::SearchIdx'; my ($tmpdir, $for_destroy) = tmpdir(); my $ibx = PublicInbox::Inbox->new({