475 lines
16 KiB
Perl
Executable file
475 lines
16 KiB
Perl
Executable file
#!/usr/bin/perl
|
|
|
|
# daemon to watch the zebraqueue and update zebra as needed
|
|
|
|
use strict;
|
|
#use warnings; FIXME - Bug 2505
|
|
BEGIN {
|
|
# find Koha's Perl modules
|
|
# test carefully before changing this
|
|
use FindBin;
|
|
eval { require "$FindBin::Bin/kohalib.pl" };
|
|
}
|
|
|
|
use POE qw(Wheel::SocketFactory Wheel::ReadWrite Filter::Stream Driver::SysRW);
|
|
use Unix::Syslog qw(:macros);
|
|
|
|
use C4::Context;
|
|
use C4::Biblio;
|
|
use C4::Search;
|
|
use C4::AuthoritiesMarc;
|
|
use XML::Simple;
|
|
use POSIX;
|
|
use utf8;
|
|
|
|
|
|
# wait periods governing connection attempts
|
|
my $min_connection_wait = 1; # start off at 1 second
|
|
my $max_connection_wait = 1024; # max about 17 minutes
|
|
|
|
# keep separate wait period for bib and authority Zebra databases
|
|
my %zoom_connection_waits = ();
|
|
|
|
my $db_connection_wait = $min_connection_wait;
|
|
|
|
# ZOOM and Z39.50 errors that are potentially
|
|
# resolvable by connecting again and retrying
|
|
# the operation
|
|
my %retriable_zoom_errors = (
|
|
10000 => 'ZOOM_ERROR_CONNECT',
|
|
10001 => 'ZOOM_ERROR_MEMORY',
|
|
10002 => 'ZOOM_ERROR_ENCODE',
|
|
10003 => 'ZOOM_ERROR_DECODE',
|
|
10004 => 'ZOOM_ERROR_CONNECTION_LOST',
|
|
10005 => 'ZOOM_ERROR_INIT',
|
|
10006 => 'ZOOM_ERROR_INTERNAL',
|
|
10007 => 'ZOOM_ERROR_TIMEOUT',
|
|
);
|
|
|
|
# structure to store updates that have
|
|
# failed and are to be retrieved. The
|
|
# structure is a hashref of hashrefs,
|
|
# e.g.,
|
|
#
|
|
# $postoned_updates->{$server}->{$record_number} = 1;
|
|
#
|
|
# If an operation is attempted and fails because
|
|
# of a retriable error (see above), the daemon
|
|
# will try several times to recover as follows:
|
|
#
|
|
# 1. close and reopen the connection to the
|
|
# Zebra server, unless the error was a timeout,
|
|
# in which case
|
|
# 2. retry the operation
|
|
#
|
|
# If, after trying this five times, the operation still
|
|
# fails, the daemon will mark the record number as
|
|
# postponed, and try to process other entries in
|
|
# zebraqueue. When an update is postponed, the
|
|
# error will be reported to syslog.
|
|
#
|
|
# If more than 100 postponed updates are
|
|
# accumulated, the daemon will assume that
|
|
# something is seriously wrong, complain loudly,
|
|
# and abort. If running under the daemon(1) command,
|
|
# this means that the daemon will respawn.
|
|
#
|
|
my $num_postponed_updates = 0;
|
|
my $postponed_updates = {};
|
|
|
|
my $max_operation_attempts = 5;
|
|
my $max_postponed_updates = 100;
|
|
|
|
# Zebra connection timeout
|
|
my $zconn_timeout = 30;
|
|
my $zconn_timeout_multiplier = 1.5;
|
|
my $max_zconn_timeout = 120;
|
|
|
|
my $ident = "Koha Zebraqueue ";
|
|
|
|
my $debug = 0;
|
|
Unix::Syslog::openlog $ident, LOG_PID, LOG_LOCAL0;
|
|
|
|
Unix::Syslog::syslog LOG_INFO, "Starting Zebraqueue log at " . scalar localtime(time) . "\n";
|
|
|
|
sub handler_start {
|
|
|
|
# Starts session. Only ever called once only really used to set an alias
|
|
# for the POE kernel
|
|
my ( $kernel, $heap, $session ) = @_[ KERNEL, HEAP, SESSION ];
|
|
|
|
my $time = localtime(time);
|
|
Unix::Syslog::syslog LOG_INFO, "$time POE Session ", $session->ID, " has started.\n";
|
|
|
|
# check status
|
|
# $kernel->yield('status_check');
|
|
$kernel->yield('sleep');
|
|
}
|
|
|
|
sub handler_sleep {
|
|
|
|
# can be used to slow down loop execution if needed
|
|
my ( $kernel, $heap, $session ) = @_[ KERNEL, HEAP, SESSION ];
|
|
use Time::HiRes qw (sleep);
|
|
Time::HiRes::sleep(0.5);
|
|
#sleep 1;
|
|
$kernel->yield('status_check');
|
|
}
|
|
|
|
sub handler_check {
|
|
# check if we need to do anything, at the moment just checks the zebraqueue, it could check other things too
|
|
my ( $kernel, $heap, $session ) = @_[ KERNEL, HEAP, SESSION ];
|
|
my $dbh = get_db_connection();
|
|
my $sth = $dbh->prepare("SELECT count(*) AS opcount FROM zebraqueue WHERE done = 0");
|
|
$sth->execute;
|
|
my $data = $sth->fetchrow_hashref();
|
|
if ($data->{'opcount'} > 0) {
|
|
Unix::Syslog::syslog LOG_INFO, "$data->{'opcount'} operations waiting to be run\n";
|
|
$sth->finish();
|
|
$dbh->commit(); # needed so that we get current state of zebraqueue next time
|
|
# we enter handler_check
|
|
$kernel->yield('do_ops');
|
|
}
|
|
else {
|
|
$sth->finish();
|
|
$dbh->commit(); # needed so that we get current state of zebraqueue next time
|
|
# we enter handler_check
|
|
$kernel->yield('sleep');
|
|
}
|
|
}
|
|
|
|
sub zebraop {
|
|
# execute operations waiting in the zebraqueue
|
|
my ( $kernel, $heap, $session ) = @_[ KERNEL, HEAP, SESSION ];
|
|
my $dbh = get_db_connection();
|
|
my $readsth = $dbh->prepare("SELECT id, biblio_auth_number, operation, server FROM zebraqueue WHERE done = 0 ORDER BY id DESC");
|
|
$readsth->execute();
|
|
Unix::Syslog::syslog LOG_INFO, "Executing zebra operations\n";
|
|
|
|
my $completed_updates = {};
|
|
ZEBRAQUEUE: while (my $data = $readsth->fetchrow_hashref()) {
|
|
warn "Inside while loop" if $debug;
|
|
|
|
my $id = $data->{'id'};
|
|
my $op = $data->{'operation'};
|
|
$op = 'recordDelete' if $op =~ /delete/i; # delete ops historically have been coded
|
|
# either delete_record or recordDelete
|
|
my $record_number = $data->{'biblio_auth_number'};
|
|
my $server = $data->{'server'};
|
|
|
|
next ZEBRAQUEUE if exists $postponed_updates->{$server}->{$record_number};
|
|
next ZEBRAQUEUE if exists $completed_updates->{$server}->{$record_number}->{$op};
|
|
|
|
my $ok = 0;
|
|
my $record;
|
|
if ($op eq 'recordDelete') {
|
|
$ok = process_delete($dbh, $server, $record_number);
|
|
}
|
|
else {
|
|
$ok = process_update($dbh, $server, $record_number, $id);
|
|
}
|
|
if ($ok == 1) {
|
|
mark_done($dbh, $record_number, $op, $server);
|
|
$completed_updates->{$server}->{$record_number}->{$op} = 1;
|
|
if ($op eq 'recordDelete') {
|
|
$completed_updates->{$server}->{$record_number}->{'specialUpdate'} = 1;
|
|
}
|
|
}
|
|
}
|
|
$readsth->finish();
|
|
$dbh->commit();
|
|
$kernel->yield('sleep');
|
|
}
|
|
|
|
sub process_delete {
|
|
my $dbh = shift;
|
|
my $server = shift;
|
|
my $record_number = shift;
|
|
|
|
my $record;
|
|
my $ok = 0;
|
|
eval {
|
|
warn "Searching for record to delete" if $debug;
|
|
# 1st read the record in zebra, we have to get it from zebra as its no longer in the db
|
|
my $Zconn = get_zebra_connection($server);
|
|
my $results = $Zconn->search_pqf( '@attr 1=Local-number '.$record_number);
|
|
$results->option(elementSetName => 'marcxml');
|
|
$record = $results->record(0)->raw();
|
|
};
|
|
if ($@) {
|
|
# this doesn't exist, so no need to wail on zebra to delete it
|
|
if ($@->code() eq 13) {
|
|
$ok = 1;
|
|
} else {
|
|
# caught a ZOOM::Exception
|
|
my $message = _format_zoom_error_message($@);
|
|
postpone_update($server, $record_number, $message);
|
|
}
|
|
} else {
|
|
# then, delete the record
|
|
warn "Deleting record" if $debug;
|
|
$ok = zebrado($record, 'recordDelete', $server, $record_number);
|
|
}
|
|
return $ok;
|
|
}
|
|
|
|
sub process_update {
|
|
my $dbh = shift;
|
|
my $server = shift;
|
|
my $record_number = shift;
|
|
my $id = shift;
|
|
|
|
my $record;
|
|
my $ok = 0;
|
|
|
|
warn "Updating record" if $debug;
|
|
# get the XML
|
|
my $marcxml;
|
|
if ($server eq "biblioserver") {
|
|
my $marc = GetMarcBiblio($record_number);
|
|
$marcxml = $marc->as_xml_record() if $marc;
|
|
}
|
|
elsif ($server eq "authorityserver") {
|
|
$marcxml = C4::AuthoritiesMarc::GetAuthorityXML($record_number);
|
|
}
|
|
# check it's XML, just in case
|
|
eval {
|
|
my $hashed = XMLin($marcxml);
|
|
}; ### is it a proper xml? broken xml may crash ZEBRA- slow but safe
|
|
## it's Broken XML-- Should not reach here-- but if it does -lets protect ZEBRA
|
|
if ($@) {
|
|
Unix::Syslog::syslog LOG_ERR, "$server record $record_number is malformed: $@";
|
|
mark_done_by_id($dbh, $id, $server);
|
|
$ok = 0;
|
|
} else {
|
|
# ok, we have everything, do the operation in zebra !
|
|
$ok = zebrado($marcxml, 'specialUpdate', $server, $record_number);
|
|
}
|
|
return $ok;
|
|
}
|
|
|
|
sub mark_done_by_id {
|
|
my $dbh = shift;
|
|
my $id = shift;
|
|
my $server = shift;
|
|
my $delsth = $dbh->prepare("UPDATE zebraqueue SET done = 1 WHERE id = ? AND server = ? AND done = 0");
|
|
$delsth->execute($id, $server);
|
|
}
|
|
|
|
sub mark_done {
|
|
my $dbh = shift;
|
|
my $record_number = shift;
|
|
my $op = shift;
|
|
my $server = shift;
|
|
|
|
my $delsth;
|
|
if ($op eq 'recordDelete') {
|
|
# if it's a deletion, we can delete every request on this biblio : in case the user
|
|
# did a modif (or item deletion) just before biblio deletion, there are some specialUpdate
|
|
# that are pending and can't succeed, as we don't have the XML anymore
|
|
# so, delete everything for this biblionumber
|
|
$delsth = $dbh->prepare_cached("UPDATE zebraqueue SET done = 1
|
|
WHERE biblio_auth_number = ?
|
|
AND server = ?
|
|
AND done = 0");
|
|
$delsth->execute($record_number, $server);
|
|
} else {
|
|
# if it's not a deletion, delete every pending specialUpdate for this biblionumber
|
|
# in case the user add biblio, then X items, before this script runs
|
|
# this avoid indexing X+1 times where just 1 is enough.
|
|
$delsth = $dbh->prepare("UPDATE zebraqueue SET done = 1
|
|
WHERE biblio_auth_number = ?
|
|
AND operation = 'specialUpdate'
|
|
AND server = ?
|
|
AND done = 0");
|
|
$delsth->execute($record_number, $server);
|
|
}
|
|
}
|
|
|
|
sub zebrado {
|
|
###Accepts a $server variable thus we can use it to update biblios, authorities or other zebra dbs
|
|
my ($record, $op, $server, $record_number) = @_;
|
|
|
|
unless ($record) {
|
|
my $message = "error updating index for $server $record $record_number: no source record";
|
|
postpone_update($server, $record_number, $message);
|
|
return 0;
|
|
}
|
|
|
|
my $attempts = 0;
|
|
my $ok = 0;
|
|
ATTEMPT: while ($attempts < $max_operation_attempts) {
|
|
$attempts++;
|
|
warn "Attempt $attempts for $op for $server $record_number" if $debug;
|
|
my $Zconn = get_zebra_connection($server);
|
|
|
|
my $Zpackage = $Zconn->package();
|
|
$Zpackage->option(action => $op);
|
|
$Zpackage->option(record => $record);
|
|
|
|
eval { $Zpackage->send("update") };
|
|
if ($@ && $@->isa("ZOOM::Exception")) {
|
|
my $message = _format_zoom_error_message($@);
|
|
my $error = $@->code();
|
|
if (exists $retriable_zoom_errors{$error}) {
|
|
warn "reattempting operation $op for $server $record_number" if $debug;
|
|
warn "last Zebra error was $message" if $debug;
|
|
$Zpackage->destroy();
|
|
|
|
if ($error == 10007 and $zconn_timeout < $max_zconn_timeout) {
|
|
# bump up connection timeout
|
|
$zconn_timeout = POSIX::ceil($zconn_timeout * $zconn_timeout_multiplier);
|
|
$zconn_timeout = $max_zconn_timeout if $zconn_timeout > $max_zconn_timeout;
|
|
Unix::Syslog::syslog LOG_INFO, "increased Zebra connection timeout to $zconn_timeout\n";
|
|
warn "increased Zebra connection timeout to $zconn_timeout" if $debug;
|
|
}
|
|
next ATTEMPT;
|
|
} else {
|
|
postpone_update($server, $record_number, $message);
|
|
}
|
|
}
|
|
# FIXME - would be more efficient to send a ES commit
|
|
# after a batch of records, rather than commiting after
|
|
# each one - Zebra handles updates relatively slowly.
|
|
eval { $Zpackage->send('commit'); };
|
|
if ($@) {
|
|
# operation succeeded, but commit
|
|
# did not - we have a problem
|
|
my $message = _format_zoom_error_message($@);
|
|
postpone_update($server, $record_number, $message);
|
|
} else {
|
|
$ok = 1;
|
|
last ATTEMPT;
|
|
}
|
|
}
|
|
|
|
unless ($ok) {
|
|
my $message = "Made $attempts attempts to index $server record $record_number without success";
|
|
postpone_update($server, $record_number, $message);
|
|
}
|
|
|
|
return $ok;
|
|
}
|
|
|
|
sub postpone_update {
|
|
my ($server, $record_number, $message) = @_;
|
|
warn $message if $debug;
|
|
$message .= "\n" unless $message =~ /\n$/;
|
|
Unix::Syslog::syslog LOG_ERR, $message;
|
|
$postponed_updates->{$server}->{$record_number} = 1;
|
|
|
|
$num_postponed_updates++;
|
|
if ($num_postponed_updates > $max_postponed_updates) {
|
|
warn "exiting, over $max_postponed_updates postponed indexing updates";
|
|
Unix::Syslog::syslog LOG_ERR, "exiting, over $max_postponed_updates postponed indexing updates";
|
|
Unix::Syslog::closelog;
|
|
exit;
|
|
}
|
|
}
|
|
|
|
sub handler_stop {
|
|
my $heap = $_[HEAP];
|
|
my $time = localtime(time);
|
|
Unix::Syslog::syslog LOG_INFO, "$time Session ", $_[SESSION]->ID, " has stopped.\n";
|
|
delete $heap->{session};
|
|
}
|
|
|
|
# get a DB connection
|
|
sub get_db_connection {
|
|
my $dbh;
|
|
|
|
$db_connection_wait = $min_connection_wait unless defined $db_connection_wait;
|
|
while (1) {
|
|
eval {
|
|
# note that C4::Context caches the
|
|
# DB handle; C4::Context->dbh() will
|
|
# check that handle first before returning
|
|
# it. If the connection is bad, it
|
|
# then tries (once) to create a new one.
|
|
$dbh = C4::Context->dbh();
|
|
};
|
|
|
|
unless ($@) {
|
|
# C4::Context->dbh dies if it cannot
|
|
# establish a connection
|
|
$db_connection_wait = $min_connection_wait;
|
|
$dbh->{AutoCommit} = 0; # do this to reduce number of
|
|
# commits to zebraqueue
|
|
return $dbh;
|
|
}
|
|
|
|
# connection failed
|
|
my $error = "failed to connect to DB: $DBI::errstr";
|
|
warn $error if $debug;
|
|
Unix::Syslog::syslog LOG_ERR, $error;
|
|
sleep $db_connection_wait;
|
|
$db_connection_wait *= 2 unless $db_connection_wait >= $max_connection_wait;
|
|
}
|
|
}
|
|
|
|
# get a Zebra connection
|
|
sub get_zebra_connection {
|
|
my $server = shift;
|
|
|
|
# start connection retry wait queue if necessary
|
|
$zoom_connection_waits{$server} = $min_connection_wait unless exists $zoom_connection_waits{$server};
|
|
|
|
# try to connect to Zebra forever until we succeed
|
|
while (1) {
|
|
# what follows assumes that C4::Context->Zconn
|
|
# makes only one attempt to create a new connection;
|
|
my $Zconn = C4::Context->Zconn($server, 0, 1, '', 'xml');
|
|
$Zconn->option('timeout' => $zconn_timeout);
|
|
|
|
# it is important to note that if the existing connection
|
|
# stored by C4::Context has an error (any type of error)
|
|
# from the last transaction, C4::Context->Zconn closes
|
|
# it and establishes a new one. Therefore, the
|
|
# following check will succeed if we have a new, good
|
|
# connection or we're using a previously established
|
|
# connection that has experienced no errors.
|
|
if ($Zconn->errcode() == 0) {
|
|
$zoom_connection_waits{$server} = $min_connection_wait;
|
|
return $Zconn;
|
|
}
|
|
|
|
# connection failed
|
|
my $error = _format_zoom_error_message($Zconn);
|
|
warn $error if $debug;
|
|
Unix::Syslog::syslog LOG_ERR, $error;
|
|
sleep $zoom_connection_waits{$server};
|
|
$zoom_connection_waits{$server} *= 2 unless $zoom_connection_waits{$server} >= $max_connection_wait;
|
|
}
|
|
}
|
|
|
|
# given a ZOOM::Exception or
|
|
# ZOOM::Connection object, generate
|
|
# a human-reaable error message
|
|
sub _format_zoom_error_message {
|
|
my $err = shift;
|
|
|
|
my $message = "";
|
|
if (ref($err) eq 'ZOOM::Connection') {
|
|
$message = $err->errmsg() . " (" . $err->diagset . " " . $err->errcode() . ") " . $err->addinfo();
|
|
} elsif (ref($err) eq 'ZOOM::Exception') {
|
|
$message = $err->message() . " (" . $err->diagset . " " . $err->code() . ") " . $err->addinfo();
|
|
}
|
|
return $message;
|
|
}
|
|
|
|
POE::Session->create(
|
|
inline_states => {
|
|
_start => \&handler_start,
|
|
sleep => \&handler_sleep,
|
|
status_check => \&handler_check,
|
|
do_ops => \&zebraop,
|
|
_stop => \&handler_stop,
|
|
},
|
|
);
|
|
|
|
# start the kernel
|
|
$poe_kernel->run();
|
|
|
|
Unix::Syslog::closelog;
|
|
|
|
exit;
|