git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jakub Narebski <jnareb@gmail.com>
To: git@vger.kernel.org
Cc: John 'Warthog9' Hawley <warthog9@eaglescrag.net>,
	John 'Warthog9' Hawley <warthog9@kernel.org>,
	Jakub Narebski <jnareb@gmail.com>
Subject: [RFC PATCH 09/10] gitweb/cache.pm - Serve stale data when waiting for filling cache (WIP)
Date: Sat, 23 Jan 2010 01:27:32 +0100	[thread overview]
Message-ID: <dba7265713bde1e403c3cbab6de81345b20f8ab8.1264198194.git.jnareb@gmail.com> (raw)
In-Reply-To: <1263432185-21334-10-git-send-email-warthog9@eaglescrag.net>
In-Reply-To: <cover.1264198194.git.jnareb@gmail.com>

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
---
This patch, as you can see, lack proper commit message: it is work in
progress.

Note that contrary to original patch by J.H. we fork unconditionally;
otherwise the process which do recalculate cache entry would be at
disadvantage compared to processes which failed race to acquire
writers lock.  Perhaps this should be made configurable, though.

Also note that currently there is no limit on how stale the data which
is being served while regenerating data can be, i.e. there is no
equivalent for $maxCacheLife.  Simplest solution (I guess) would be to
just delete cache entry if it is too stale upfront.

 gitweb/cache.pm                 |   32 +++++++++++++++++++++++++++-----
 t/t9503/test_cache_interface.pl |   37 +++++++++++++++++++++++++++++++++++++
 2 files changed, 64 insertions(+), 5 deletions(-)

diff --git a/gitweb/cache.pm b/gitweb/cache.pm
index f514ee9..1adf467 100644
--- a/gitweb/cache.pm
+++ b/gitweb/cache.pm
@@ -426,14 +426,36 @@ sub compute {
 	_Make_Path($lockfile);
 	open my $lock_fh, '+>', $lockfile;
 	#	or die "Can't open lockfile '$lockfile': $!";
+
+	# try to retrieve stale data
+	$data = $self->restore($self->get_namespace(), $p_key);
+
 	if (my $lock_state = flock($lock_fh, LOCK_EX | LOCK_NB)) {
 		# acquired writers lock
-		$data = $p_coderef->($self, $p_key);
-		$self->set($p_key, $data);
+		my $pid = fork() if $data;
+		if (!defined $pid || $pid) {
+			# parent, or didn't fork
+			$data = $p_coderef->($self, $p_key);
+			$self->set($p_key, $data);
+
+			if ($pid) {
+				# wait for child (which would print) and exit
+				waitpid $pid, 0;
+				exit 0;
+			} else {
+				# there is no child, or was no $data to serve in background
+				;
+			}
+		} else {
+			# child to serve $data
+			;
+		}
 	} else {
-		# get readers lock
-		flock($lock_fh, LOCK_SH);
-		$data = $self->restore($self->get_namespace(), $p_key);
+		if (!defined $data) {
+			# get readers lock if there is no stale data to serve
+			flock($lock_fh, LOCK_SH);
+			$data = $self->restore($self->get_namespace(), $p_key);
+		}
 	}
 	close $lock_fh;
 	return $data;
diff --git a/t/t9503/test_cache_interface.pl b/t/t9503/test_cache_interface.pl
index 43b806d..788e8f2 100755
--- a/t/t9503/test_cache_interface.pl
+++ b/t/t9503/test_cache_interface.pl
@@ -154,5 +154,42 @@ is($cache->get_expires_in(), 0,        '"expires in" is set to now (0)');
 $cache->set($key, $value);
 ok(!defined($cache->get($key)),        'cache is expired');
 
+# Test that cache returns stale data in existing but expired cache situation
+# (probably should be run only if GIT_TEST_LONG)
+$cache->set_expires_min(0);
+$cache->set_expires_max(0);
+my $stale_value = 'Stale Value';
+my $child_data = '';
+$cache->set($key, $stale_value);
+$call_count = 0;
+$pid = open $kid_fh, '-|';
+SKIP: {
+	skip "cannot fork: $!", 4
+		unless defined $pid;
+
+	my $data = $cache->compute($key, \&get_value_slow);
+
+	if ($pid) {
+		$child_data = <$kid_fh>;
+		chomp $child_data;
+
+		waitpid $pid, 0;
+		close $kid_fh;
+	} else {
+		print "$data\n";
+		exit 0;
+	}
+
+	is($data,       $stale_value, 'stale data in parent when expired');
+	is($child_data, $stale_value, 'stale data in child  when expired');
+
+	# never expire
+	$cache->set_expires_min(-1);
+	$cache->set_expires_max(-1);
+	is($cache->get($key), $value, 'value got set correctly');
+}
+$cache->set_expires_min(0);
+$cache->set_expires_max(0);
+
 
 done_testing();
-- 
1.6.6.1

  parent reply	other threads:[~2010-01-23  0:28 UTC|newest]

Thread overview: 60+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-01-14  1:22 [PATCH 0/9] Gitweb caching v5 John 'Warthog9' Hawley
2010-01-14  1:22 ` [PATCH 1/9] gitweb: Load checking John 'Warthog9' Hawley
2010-01-14  1:22   ` [PATCH 2/9] gitweb: change die_error to take "extra" argument for extended die information John 'Warthog9' Hawley
2010-01-14  1:22     ` [PATCH 3/9] gitweb: Add option to force version match John 'Warthog9' Hawley
2010-01-14  1:23       ` [PATCH 4/9] gitweb: Makefile improvements John 'Warthog9' Hawley
2010-01-14  1:23         ` [PATCH 5/9] gitweb: add a get function to compliment print_local_time John 'Warthog9' Hawley
2010-01-14  1:23           ` [PATCH 6/9] gitweb: add a get function to compliment print_sort_th John 'Warthog9' Hawley
2010-01-14  1:23             ` [PATCH 7/9] gitweb: cleanup error message produced by undefined $site_header John 'Warthog9' Hawley
2010-01-14  1:23               ` [PATCH 8/9] gitweb: Convert output to using indirect file handle John 'Warthog9' Hawley
2010-01-14  1:23                 ` [PATCH 9/9] gitweb: File based caching layer (from git.kernel.org) John 'Warthog9' Hawley
2010-01-16  2:48                   ` Jakub Narebski
2010-01-23  0:27                   ` [RFC PATCH 00/10] gitweb: Simple file based output caching Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 02/10] gitweb: href(..., -path_info => 0|1) Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 03/10] gitweb/cache.pm - Very simple file based caching Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 04/10] gitweb/cache.pm - Stat-based cache expiration Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 05/10] gitweb: Use Cache::Cache compatibile (get, set) output caching (WIP) Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 06/10] gitweb/cache.pm - Adaptive cache expiration time (WIP) Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 07/10] gitweb: Use CHI compatibile (compute method) caching (WIP) Jakub Narebski
2010-01-23  0:27                     ` [RFC PATCH 08/10] gitweb/cache.pm - Use locking to avoid 'stampeding herd' problem (WIP) Jakub Narebski
2010-01-23  0:27                     ` Jakub Narebski [this message]
2010-01-23  0:27                     ` [RFC PATCH 10/10] gitweb: Show appropriate "Generating..." page when regenerating cache (WIP) Jakub Narebski
2010-01-24 22:24                       ` Petr Baudis
2010-01-25  0:03                         ` Jakub Narebski
2010-01-25  1:17                           ` Jakub Narebski
2010-01-25 11:46                         ` Jakub Narebski
2010-01-25 13:02                           ` Petr Baudis
2010-01-25 13:48                             ` Jakub Narebski
2010-01-25 13:56                               ` Petr Baudis
2010-01-25 20:32                                 ` J.H.
2010-01-26  1:49                                   ` Jakub Narebski
2010-01-28 17:39                                   ` Petr Baudis
2010-01-31 11:58                                     ` Jakub Narebski
2010-01-25 20:58                                 ` Jakub Narebski
2010-01-25 20:41                               ` J.H.
2010-01-26  2:30                                 ` Jakub Narebski
2010-01-23 19:55                     ` [RFC PATCH 00/10] gitweb: Simple file based output caching J.H.
2010-01-24 13:54                     ` [RFC PATCH 11/10] gitweb: Ajax-y "Generating..." page when regenerating cache (WIP) Jakub Narebski
2010-02-06  0:51                     ` [RFC PATCH 00/10] gitweb: Simple file based output caching J.H.
2010-02-06 23:56                       ` Jakub Narebski
2010-02-07 12:35                         ` Jakub Narebski
     [not found]                   ` <0dd15cb3f18e2a26fc834fd3b071e6d3ecc00557.1264198194.git.jnareb@gmail.com>
2010-01-23  0:48                     ` [RFC PATCH 01/10] gitweb: Print to explicit filehandle (preparing for caching) Jakub Narebski
2010-02-07 21:32                     ` Jakub Narebski
2010-01-16  0:43                 ` [PATCH 8/9] gitweb: Convert output to using indirect file handle Jakub Narebski
2010-01-16  0:58                   ` Junio C Hamano
2010-01-16  1:14                     ` Jakub Narebski
2010-01-16  1:41                       ` Junio C Hamano
2010-01-24 22:14                   ` Petr Baudis
2010-01-25  1:47                     ` Jakub Narebski
2010-01-25 20:48                       ` J.H.
2010-01-25 21:48                         ` Jakub Narebski
2010-01-15 23:49               ` [PATCH 7/9] gitweb: cleanup error message produced by undefined $site_header Jakub Narebski
2010-01-23 11:13           ` [PATCH 5/9] gitweb: add a get function to compliment print_local_time Jakub Narebski
2010-01-15 23:36       ` [PATCH 3/9] gitweb: Add option to force version match Jakub Narebski
2010-01-24 21:59       ` Petr Baudis
2010-01-24 23:17         ` Jakub Narebski
2010-01-15 22:40     ` [PATCH 2/9] gitweb: change die_error to take "extra" argument for extended die information Jakub Narebski
2010-01-15 22:30   ` [PATCH 1/9] gitweb: Load checking Jakub Narebski
2010-01-15  1:40 ` [PATCH 0/9] Gitweb caching v5 Jakub Narebski
2010-01-15  4:29   ` J.H.
2010-01-15 10:28     ` Jakub Narebski

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

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

  git send-email \
    --in-reply-to=dba7265713bde1e403c3cbab6de81345b20f8ab8.1264198194.git.jnareb@gmail.com \
    --to=jnareb@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=warthog9@eaglescrag.net \
    --cc=warthog9@kernel.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.
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 NNTP newsgroup(s).