X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FExtSearchIdx.pm;h=6b29789a2ed89d33d12e654d6a95f9bbe65cc8fa;hb=28d5a8d647e3ab56cc5570af0d6f3ccf75dc91f2;hp=1c2a97588e3432960d87ad50390f20bf35317adf;hpb=6896585267c269cb344b01af102f6f73602548ad;p=public-inbox.git diff --git a/lib/PublicInbox/ExtSearchIdx.pm b/lib/PublicInbox/ExtSearchIdx.pm index 1c2a9758..6b29789a 100644 --- a/lib/PublicInbox/ExtSearchIdx.pm +++ b/lib/PublicInbox/ExtSearchIdx.pm @@ -21,6 +21,7 @@ use Carp qw(croak carp); use Sys::Hostname qw(hostname); use POSIX qw(strftime); use File::Glob qw(bsd_glob GLOB_NOSORT); +use PublicInbox::MultiGit; use PublicInbox::Search; use PublicInbox::SearchIdx qw(prepare_stack is_ancestor is_bad_blob); use PublicInbox::OverIdx; @@ -44,7 +45,8 @@ sub new { topdir => $dir, creat => $opt->{creat}, ibx_map => {}, # (newsgroup//inboxdir) => $ibx - ibx_cfg => [], # by config section order + ibx_active => [], # by config section order + ibx_known => [], # by config section order indexlevel => $l, transact_bytes => 0, total_bytes => 0, @@ -52,7 +54,8 @@ sub new { parallel => 1, lock_path => "$dir/ei.lock", }, __PACKAGE__; - $self->{shards} = $self->count_shards || nproc_shards($opt->{creat}); + $self->{shards} = $self->count_shards || + nproc_shards({ nproc => $opt->{jobs} }); my $oidx = PublicInbox::OverIdx->new("$self->{xpfx}/over.sqlite3"); $self->{-no_fsync} = $oidx->{-no_fsync} = 1 if !$opt->{fsync}; $self->{oidx} = $oidx; @@ -60,23 +63,41 @@ sub new { } sub attach_inbox { - my ($self, $ibx) = @_; + my ($self, $ibx, $types) = @_; $self->{ibx_map}->{$ibx->eidx_key} //= do { - delete $self->{-ibx_ary}; # invalidate cache - push @{$self->{ibx_cfg}}, $ibx; + delete $self->{-ibx_ary_known}; # invalidate cache + delete $self->{-ibx_ary_active}; # invalidate cache + $types //= [ qw(active known) ]; + for my $t (@$types) { + push @{$self->{"ibx_$t"}}, $ibx; + } $ibx; } } sub _ibx_attach { # each_inbox callback - my ($ibx, $self) = @_; - attach_inbox($self, $ibx); + my ($ibx, $self, $types) = @_; + attach_inbox($self, $ibx, $types); } sub attach_config { - my ($self, $cfg) = @_; + my ($self, $cfg, $ibxs) = @_; $self->{cfg} = $cfg; - $cfg->each_inbox(\&_ibx_attach, $self); + my $types; + if ($ibxs) { + for my $ibx (@$ibxs) { + $self->{ibx_map}->{$ibx->eidx_key} //= do { + push @{$self->{ibx_active}}, $ibx; + push @{$self->{ibx_known}}, $ibx; + } + } + # invalidate cache + delete $self->{-ibx_ary_known}; + delete $self->{-ibx_ary_active}; + $types = [ 'known' ]; + } + $types //= [ qw(known active) ]; + $cfg->each_inbox(\&_ibx_attach, $self, $types); } sub check_batch_limit ($) { @@ -89,6 +110,25 @@ sub check_batch_limit ($) { ${$req->{need_checkpoint}} = 1 if $n >= $self->{batch_bytes}; } +sub apply_boost ($$) { + my ($req, $smsg) = @_; + my $id2pos = $req->{id2pos}; # index in ibx_sorted + my $xr3 = $req->{self}->{oidx}->get_xref3($smsg->{num}, 1); + @$xr3 = sort { + $id2pos->{$a->[0]} <=> $id2pos->{$b->[0]} + || + $a->[1] <=> $b->[1] # break ties with {xnum} + } @$xr3; + my $top_blob = unpack('H*', $xr3->[0]->[2]); + my $new_smsg = $req->{new_smsg}; + return if $top_blob ne $new_smsg->{blob}; # loser + + # replace the old smsg with the more boosted one + $new_smsg->{num} = $smsg->{num}; + $new_smsg->populate($req->{eml}, $req); + $req->{self}->{oidx}->add_overview($req->{eml}, $new_smsg); +} + sub do_xpost ($$) { my ($req, $smsg) = @_; my $self = $req->{self}; @@ -102,7 +142,7 @@ sub do_xpost ($$) { my $xnum = $req->{xnum}; $self->{oidx}->add_xref3($docid, $xnum, $oid, $eidx_key); $idx->ipc_do('add_eidx_info', $docid, $eidx_key, $eml); - check_batch_limit($req); + apply_boost($req, $smsg) if $req->{boost_in_use}; } else { # 'd' my $rm_eidx_info; my $nr = $self->{oidx}->remove_xref3($docid, $oid, $eidx_key, @@ -304,20 +344,18 @@ sub _sync_inbox ($$$) { sub gc_unref_doc ($$$$) { my ($self, $ibx_id, $eidx_key, $docid) = @_; - my $dbh = $self->{oidx}->dbh; - + my $remain = 0; # for debug/info purposes, oids may no longer be accessible + my $dbh = $self->{oidx}->dbh; my $sth = $dbh->prepare_cached(<<'', undef, 1); SELECT oidbin FROM xref3 WHERE docid = ? AND ibx_id = ? $sth->execute($docid, $ibx_id); my @oid = map { unpack('H*', $_->[0]) } @{$sth->fetchall_arrayref}; - - $dbh->prepare_cached(<<'')->execute($docid, $ibx_id); -DELETE FROM xref3 WHERE docid = ? AND ibx_id = ? - - my $remain = $self->{oidx}->get_xref3($docid); - if (scalar(@$remain)) { + for my $oid (@oid) { + $remain += $self->{oidx}->remove_xref3($docid, $oid, $eidx_key); + } + if ($remain) { $self->{oidx}->eidxq_add($docid); # enqueue for reindex for my $oid (@oid) { warn "I: unref #$docid $eidx_key $oid\n"; @@ -382,6 +420,11 @@ DELETE FROM xref3 WHERE docid NOT IN (SELECT num FROM over) warn "I: eliminated $nr stale xref3 entries\n" if $nr != 0; + # fixup from old bugs: + $nr = $dbh->do(<<''); +DELETE FROM over WHERE num NOT IN (SELECT docid FROM xref3) + + warn "I: eliminated $nr stale over entries\n" if $nr != 0; done($self); } @@ -389,7 +432,8 @@ sub _ibx_for ($$$) { my ($self, $sync, $smsg) = @_; my $ibx_id = delete($smsg->{ibx_id}) // die '{ibx_id} unset'; my $pos = $sync->{id2pos}->{$ibx_id} // die "$ibx_id no pos"; - $self->{-ibx_ary}->[$pos] // die "BUG: ibx for $smsg->{blob} not mapped" + $self->{-ibx_ary_known}->[$pos] // + die "BUG: ibx for $smsg->{blob} not mapped" } sub _fd_constrained ($) { @@ -403,7 +447,8 @@ sub _fd_constrained ($) { chomp($soft = `sh -c 'ulimit -n'`); } if (defined($soft)) { - my $want = scalar(@{$self->{-ibx_ary}}) + 64; # estimate + # $want is an estimate + my $want = scalar(@{$self->{ibx_active}}) + 64; my $ret = $want > $soft; if ($ret) { warn <{-ibx_ary} //= do { +sub ibx_sorted ($$) { + my ($self, $type) = @_; + $self->{"-ibx_ary_$type"} //= do { # highest boost first, stable for config-ordering tiebreaker use sort 'stable'; [ sort { ($b->{boost} // 0) <=> ($a->{boost} // 0) - } @{$self->{ibx_cfg}} ]; + } @{$self->{'ibx_'.$type} // die "BUG: $type unknown"} ]; } } +sub prep_id2pos ($) { + my ($self) = @_; + my %id2pos; + my $pos = 0; + $id2pos{$_->{-ibx_id}} = $pos++ for (@{ibx_sorted($self, 'known')}); + \%id2pos; +} + sub eidxq_process ($$) { # for reindexing my ($self, $sync) = @_; @@ -647,12 +700,7 @@ sub eidxq_process ($$) { # for reindexing my $max = $dbh->selectrow_array('SELECT MAX(docid) FROM eidxq'); $pr->("Xapian indexing $min..$max (total=$tot)\n"); } - $sync->{id2pos} //= do { - my %id2pos; - my $pos = 0; - $id2pos{$_->{-ibx_id}} = $pos++ for (@{ibx_sorted($self)}); - \%id2pos; - }; + $sync->{id2pos} //= prep_id2pos($self); my ($del, $iter); restart: $del = $dbh->prepare('DELETE FROM eidxq WHERE docid = ?'); @@ -784,16 +832,16 @@ ORDER BY docid,xnum ASC LIMIT 10000 $fetching = $min = $docid; my $smsg = $ibx->over->get_art($xnum); - my $oidhex = unpack('H*', $oidbin); my $err; if (!$smsg) { $err = 'stale'; - } elsif ($smsg->{blob} ne $oidhex) { + } elsif (pack('H*', $smsg->{blob}) ne $oidbin) { $err = "mismatch (!= $smsg->{blob})"; } else { next; # likely, all good } # current_info already has eidx_key + my $oidhex = unpack('H*', $oidbin); warn "$xnum:$oidhex (#$docid): $err\n"; my $del = $self->{oidx}->dbh->prepare_cached(<<''); DELETE FROM xref3 WHERE ibx_id = ? AND xnum = ? AND oidbin = ? @@ -841,7 +889,7 @@ sub eidx_reindex { warn "E: aborting --reindex\n"; return; } - for my $ibx (@{ibx_sorted($self)}) { + for my $ibx (@{ibx_sorted($self, 'active')}) { _reindex_inbox($self, $sync, $ibx); last if $sync->{quit}; } @@ -896,7 +944,10 @@ sub eidx_dedupe ($$$) { my ($iter, $cur_mid); my $min_id = 0; my $idx = 0; - local $sync->{-regen_fmt} = "dedupe %u/".$self->{oidx}->max."\n"; + my ($max_id) = $self->{oidx}->dbh->selectrow_array(<{-regen_fmt} = "dedupe %u/$max_id\n"; # note: we could write this query more intelligently, # but that causes lock contention with read-only processes @@ -984,9 +1035,15 @@ sub eidx_sync { # main entry point local $SIG{QUIT} = $quit; local $SIG{INT} = $quit; local $SIG{TERM} = $quit; - for my $ibx (@{ibx_sorted($self)}) { + for my $ibx (@{ibx_sorted($self, 'known')}) { $ibx->{-ibx_id} //= $self->{oidx}->ibx_id($ibx->eidx_key); } + + if (scalar(grep { defined($_->{boost}) } @{$self->{ibx_known}})) { + $sync->{id2pos} //= prep_id2pos($self); + $sync->{boost_in_use} = 1; + } + if (my $msgids = delete($opt->{dedupe})) { local $sync->{checkpoint_unlocks} = 1; eidx_dedupe($self, $sync, $msgids); @@ -998,7 +1055,7 @@ sub eidx_sync { # main entry point # don't use $_ here, it'll get clobbered by reindex_checkpoint if ($opt->{scan} // 1) { - for my $ibx (@{ibx_sorted($self)}) { + for my $ibx (@{ibx_sorted($self, 'active')}) { last if $sync->{quit}; sync_inbox($self, $sync, $ibx); } @@ -1077,88 +1134,60 @@ sub idx_init { # similar to V2Writable $self->git->cleanup; my $mode = 0644; - my $ALL = $self->git->{git_dir}; # ALL.git - my $old = -d $ALL; + my $ALL = $self->git->{git_dir}; # topdir/ALL.git + my ($has_new, $alt, $seen); if ($opt->{-private}) { # LeiStore + my $local = "$self->{topdir}/local"; # lei/store + $self->{mg} //= PublicInbox::MultiGit->new($self->{topdir}, + 'ALL.git', 'local'); $mode = 0600; - if (!$old) { - umask 077; # don't bother restoring + unless (-d $ALL) { + umask 077; # don't bother restoring for lei PublicInbox::Import::init_bare($ALL); $self->git->qx(qw(config core.sharedRepository 0600)); } - } else { - PublicInbox::Import::init_bare($ALL) unless $old; - } - my $info_dir = "$ALL/objects/info"; - my $alt = "$info_dir/alternates"; - my (@old, @new, %seen); # seen: st_dev + st_ino - if (-e $alt) { - open(my $fh, '<', $alt) or die "open $alt: $!"; - $mode = (stat($fh))[2] & 07777; - while (my $line = <$fh>) { - chomp(my $d = $line); - - # expand relative path (/local/ stuff) - substr($d, 0, 3) eq '../' and - $d = "$ALL/objects/$d"; - if (my @st = stat($d)) { - next if $seen{"$st[0]\0$st[1]"}++; - } else { - warn "W: stat($d) failed (from $alt): $!\n"; - next if $opt->{-idx_gc}; - } - push @old, $line; - } + ($alt, $seen) = $self->{mg}->read_alternates(\$mode); + $has_new = $self->{mg}->merge_epochs($alt, $seen); + } else { # extindex has no epochs + $self->{mg} //= PublicInbox::MultiGit->new($self->{topdir}, + 'ALL.git'); + ($alt, $seen) = $self->{mg}->read_alternates(\$mode, + $opt->{-idx_gc}); + PublicInbox::Import::init_bare($ALL); } - # for LeiStore, and possibly some mirror-only state - if (opendir(my $dh, my $local = "$self->{topdir}/local")) { - # highest numbered epoch first - for my $n (sort { $b <=> $a } map { substr($_, 0, -4) + 0 } - grep(/\A[0-9]+\.git\z/, readdir($dh))) { - my $d = "$local/$n.git/objects"; # absolute path - if (my @st = stat($d)) { - next if $seen{"$st[0]\0$st[1]"}++; - # favor relative paths for rename-friendliness - push @new, "../../local/$n.git/objects\n"; - } else { - warn "W: stat($d) failed: $!\n"; - } - } - } # git-multi-pack-index(1) can speed up "git cat-file" startup slightly - my $dh; my $git_midx = 0; my $pd = "$ALL/objects/pack"; - if (!mkdir($pd) && $!{EEXIST} && opendir($dh, $pd)) { - # drop stale symlinks + if (opendir(my $dh, $pd)) { # drop stale symlinks while (defined(my $dn = readdir($dh))) { if ($dn =~ /\.(?:idx|pack|promisor|bitmap|rev)\z/) { my $f = "$pd/$dn"; unlink($f) if -l $f && !-e $f; } } - undef $dh; + } elsif ($!{ENOENT}) { + mkdir($pd) or die "mkdir($pd): $!"; + } else { + die "opendir($pd): $!"; } - for my $ibx (@{ibx_sorted($self)}) { + my $new = ''; + for my $ibx (@{ibx_sorted($self, 'active')}) { # create symlinks for multi-pack-index $git_midx += symlink_packs($ibx, $pd); # add new lines to our alternates file - my $line = $ibx->git->{git_dir} . "/objects\n"; - chomp(my $d = $line); + my $d = $ibx->git->{git_dir} . '/objects'; + next if exists $alt->{$d}; if (my @st = stat($d)) { - next if $seen{"$st[0]\0$st[1]"}++; + next if $seen->{"$st[0]\0$st[1]"}++; } else { warn "W: stat($d) failed (from $ibx->{inboxdir}): $!\n"; next if $opt->{-idx_gc}; } - push @new, $line; - } - if (scalar @new) { - push @old, @new; - my $o = \@old; - PublicInbox::V2Writable::write_alternates($info_dir, $mode, $o); + $new .= "$d\n"; } + ($has_new || $new ne '') and + $self->{mg}->write_alternates($mode, $alt, $new); $git_midx and $self->with_umask(sub { my @cmd = ('multi-pack-index'); push @cmd, '--no-progress' if ($opt->{quiet}//0) > 1; @@ -1170,7 +1199,7 @@ sub idx_init { # similar to V2Writable $self->with_umask(\&_idx_init, $self, $opt); $self->{oidx}->begin_lazy; $self->{oidx}->eidx_prep; - $self->{midx}->create_xdb if @new; + $self->{midx}->create_xdb if $new ne ''; } sub _watch_commit { # PublicInbox::DS::add_timer callback @@ -1205,8 +1234,10 @@ sub eidx_reload { # -extindex --watch SIGHUP handler my $pr = $self->{-watch_sync}->{-opt}->{-progress}; $pr->('reloading ...') if $pr; delete $self->{-resync_queue}; - delete $self->{-ibx_ary}; - $self->{ibx_cfg} = []; + delete $self->{-ibx_ary_known}; + delete $self->{-ibx_ary_active}; + $self->{ibx_known} = []; + $self->{ibx_active} = []; %{$self->{ibx_map}} = (); delete $self->{-watch_sync}->{id2pos}; my $cfg = PublicInbox::Config->new; @@ -1220,7 +1251,7 @@ sub eidx_reload { # -extindex --watch SIGHUP handler sub eidx_resync_start ($) { # -extindex --watch SIGUSR1 handler my ($self) = @_; - $self->{-resync_queue} //= [ @{ibx_sorted($self)} ]; + $self->{-resync_queue} //= [ @{ibx_sorted($self, 'active')} ]; PublicInbox::DS::requeue($self); # trigger our ->event_step } @@ -1241,7 +1272,7 @@ sub event_step { # PublicInbox::DS::requeue callback sub eidx_watch { # public-inbox-extindex --watch main loop my ($self, $opt) = @_; - local %SIG = %SIG; + local @SIG{keys %SIG} = values %SIG; for my $sig (qw(HUP USR1 TSTP QUIT INT TERM)) { $SIG{$sig} = sub { warn "SIG$sig ignored while scanning\n" }; } @@ -1251,9 +1282,11 @@ sub eidx_watch { # public-inbox-extindex --watch main loop require PublicInbox::Sigfd; my $idler = PublicInbox::InboxIdle->new($self->{cfg}); if (!$self->{cfg}) { - $idler->watch_inbox($_) for (@{ibx_sorted($self)}); + $idler->watch_inbox($_) for (@{ibx_sorted($self, 'active')}); + } + for my $ibx (@{ibx_sorted($self, 'active')}) { + $ibx->subscribe_unlock(__PACKAGE__, $self) } - $_->subscribe_unlock(__PACKAGE__, $self) for (@{ibx_sorted($self)}); my $pr = $opt->{-progress}; $pr->("performing initial scan ...\n") if $pr; my $sync = eidx_sync($self, $opt); # initial sync @@ -1274,7 +1307,7 @@ sub eidx_watch { # public-inbox-extindex --watch main loop $sig->{QUIT} = $sig->{INT} = $sig->{TERM} = $quit; my $sigfd = PublicInbox::Sigfd->new($sig, $PublicInbox::Syscall::SFD_NONBLOCK); - %SIG = (%SIG, %$sig) if !$sigfd; + @SIG{keys %$sig} = values(%$sig) if !$sigfd; local $self->{-watch_sync} = $sync; # for ->on_inbox_unlock if (!$sigfd) { # wake up every second to accept signals if we don't