about summary refs log tree commit homepage
path: root/lib/PublicInbox/SearchIdxPart.pm
diff options
context:
space:
mode:
authorEric Wong (Contractor, The Linux Foundation) <e@80x24.org>2018-04-02 00:04:52 +0000
committerEric Wong (Contractor, The Linux Foundation) <e@80x24.org>2018-04-02 00:05:39 +0000
commit35ff6bb106909b1c1232666a9792156dfa398ea8 (patch)
tree000f656d3daf3a077fbfa02b5853523d66a89329 /lib/PublicInbox/SearchIdxPart.pm
parent7503aeb540af5afd5cb1b554b3c29f35f5fc918d (diff)
downloadpublic-inbox-35ff6bb106909b1c1232666a9792156dfa398ea8.tar.gz
This ought to provide better performance and scalability
which is less dependent on inbox size.  Xapian does not
seem optimized for some queries used by the WWW homepage,
Atom feeds, XOVER and NEWNEWS NNTP commands.

This can actually make Xapian optional for NNTP usage,
and allow more functionality to work without Xapian
installed.

Indexing performance was extremely bad at first, but
DBI::Profile helped me optimize away problematic queries.
Diffstat (limited to 'lib/PublicInbox/SearchIdxPart.pm')
-rw-r--r--lib/PublicInbox/SearchIdxPart.pm16
1 files changed, 7 insertions, 9 deletions
diff --git a/lib/PublicInbox/SearchIdxPart.pm b/lib/PublicInbox/SearchIdxPart.pm
index 82f5c1bc..e5766a82 100644
--- a/lib/PublicInbox/SearchIdxPart.pm
+++ b/lib/PublicInbox/SearchIdxPart.pm
@@ -6,12 +6,12 @@ use warnings;
 use base qw(PublicInbox::SearchIdx);
 
 sub new {
-        my ($class, $v2writable, $part, $skel) = @_;
+        my ($class, $v2writable, $part) = @_;
         my $self = $class->SUPER::new($v2writable->{-inbox}, 1, $part);
-        $self->{skeleton} = $skel;
-        # create the DB:
+        # create the DB before forking:
         $self->_xdb_acquire;
         $self->_xdb_release;
+        $self->{over} = $v2writable->{over};
         $self->spawn_worker($v2writable, $part) if $v2writable->{parallel};
         $self;
 }
@@ -27,7 +27,7 @@ sub spawn_worker {
         if ($pid == 0) {
                 $v2writable->atfork_child;
                 $v2writable = undef;
-                close $w;
+                close $w or die "failed to close: $!";
 
                 # F_SETPIPE_SZ = 1031 on Linux; increasing the pipe size here
                 # speeds V2Writable batch imports across 8 cores by nearly 20%
@@ -40,7 +40,7 @@ sub spawn_worker {
         }
         $self->{pid} = $pid;
         $self->{w} = $w;
-        close $r;
+        close $r or die "failed to close: $!";
 }
 
 sub partition_worker_loop ($$$) {
@@ -50,13 +50,12 @@ sub partition_worker_loop ($$$) {
         while (my $line = $r->getline) {
                 if ($line eq "commit\n") {
                         $self->commit_txn_lazy;
-                        $self->{skeleton}->remote_commit;
                 } elsif ($line eq "close\n") {
                         $self->_xdb_release;
                 } elsif ($line eq "barrier\n") {
                         $self->commit_txn_lazy;
-                        print { $self->{skeleton}->{w} } "barrier $part\n" or
-                                        die "write failed to skeleton: $!\n";
+                        print { $self->{over}->{w} } "barrier $part\n" or
+                                        die "write failed to overview $!\n";
                 } elsif ($line =~ /\AD ([a-f0-9]{40,}) (.+)\n\z/s) {
                         my ($oid, $mid) = ($1, $2);
                         $self->begin_txn_lazy;
@@ -101,7 +100,6 @@ sub remote_barrier {
                 $w->flush or die "failed to flush: $!";
         } else {
                 $self->commit_txn_lazy;
-                $self->{skeleton}->remote_commit;
         }
 }