Beefy Boxes and Bandwidth Generously Provided by pair Networks
Syntactic Confectionery Delight
 
PerlMonks  

BerkeleyDB blowing up on multiple, concurrent processes

by m0r0n (Initiate)
on Jun 14, 2006 at 19:11 UTC ( [id://555331]=perlquestion: print w/replies, xml ) Need Help??

m0r0n has asked for the wisdom of the Perl Monks concerning the following question:

Dear Monks,

I have a program that's meant to stress the BerkeleyDB::Hash module to see how well it can perform with multiple, concurrent readers/writers. I am new to Perl _and_ BerkeleyDB, so go easy on me ;-)

Running the program below mostly yields the following errors:

db_put failed: DB_RUNRECOVERY: Fatal error, run database recovery ( / +Successful return: 0) main::spawn_child('BerkeleyDB::Env=ARRAY(0x82e5248)','BerkeleyDB:: +Hash=ARRAY(0x80639b4)') called at douji.pl line 115 main::spawn_child('BerkeleyDB::Env=ARRAY(0x82e5248)','BerkeleyDB:: +Hash=ARRAY(0x80639b4)') called at douji.pl line 115

One instance of the above for each child process. Sometimes I get some errors like this, too:

db_put failed: Cannot allocate memory (Cannot allocate memory / reallo +c: Cannot allocate memory: 4294967190) main::spawn_child('BerkeleyDB::Env=ARRAY(0x82e5248)','Berkeley +DB::Hash=ARRAY(0x80639b4)') called at douji.pl line 115

Running db_recover does not allow it to then finish correctly the next time, however.

Does anyone have any ideas what I'm doing wrong? I was under the impression that using DB_INIT_CDB was enough to protect against concurrent writes, but do I need transactions instead? I tried using BerkeleyDB's transaction stuff earlier and that failed with the same error messages.

TIA :)

P.S. I am using BerkeleyDB CPAN module version 0.28 on RedHat 9 using the db4 4.0.14 packages.

#!/usr/bin/env perl # vim:set ts=4 sw=4 ai: use strict; use warnings; use Carp; use POSIX ":sys_wait_h"; use Data::Dumper; use lib '/junk/lib'; use BerkeleyDB; use BerkeleyDB::Hash; use Time::HiRes qw(gettimeofday tv_interval); use constant RUNS => 1000000; use constant DBPATH => '/junk/test.db'; use constant NUM_CHILDREN => 10; $|++; # print backtrace info sub bt { croak "DIED ($.)\n"; } # generate random integer sub rand_num { 4278190080 + int(rand(1048576)); } # read an entry from a BDB sub read_db { my ($db, $key) = @_; my $val; $db->db_get($key, $val); return (defined($val)) ? eval($val) : {}; } # increment an entry in a BDB, inserting if not there sub write_db { my ($db, $key, $stuff) = @_; my $val = read_db($db, $key); $val->{timestamp} = time; $val->{events}++; $val->{stuff}++ if ($stuff); my $str = Dumper(\$val) or croak "failed to serialize row data: $! +\n"; return $db->db_put($key, $str); } # fork off a worker process that will mutate the BDB independently of # the others sub spawn_child { my ($env, $db) = @_; my $pid = fork; croak "Could not fork: '$!'\n" unless (defined($pid)); return unless ($pid == 0); # child process starts here my ($start, $finish, $key, $ip, $rv, $val, $reads, $writes, $delet +es); print "Staring worker $$...\n"; # perform a number of operations using BDB transactions $start = [ gettimeofday ]; for (1..RUNS) { $key = int(rand(101)); $ip = rand_num(); if ($key <= 49) { # do random update/insert $rv = write_db($db, $ip, ($key % 2 == 0)); croak "db_put failed: $rv ($! / $BerkeleyDB::Error)\n" unless ($rv == 0); $reads++; $writes++; } elsif ($key <= 98) { # do random read read_db($db, $ip); $reads++; } else { # do random delete $db->db_del($key); $deletes++; } } $finish = [ gettimeofday ]; # print summary info for timing my $sum = ($reads || 0) + ($writes || 0) + ($deletes || 0); my $elapsed = tv_interval($start, $finish); printf "$$: %d reads, %d writes, %d deletes (%d total) in %0.2f se +conds: %0.2f ops/sec\n", ($reads || 0), ($writes || 0), ($deletes || 0), $sum, $elapsed, $sum / $elapsed; exit 0; } # install signal handlers $SIG{__WARN__} = \&bt; # create BDB environment so we can enable transactions my $env = new BerkeleyDB::Env( -Cachesize => 4 * 1024 * 1024, -Flags => (DB_CREATE() | DB_INIT_CDB() | DB_INIT_MPOOL()), -SetFlags => DB_CDB_ALLDB()) or croak "Failed to create DB env: '$!' ($BerkeleyDB::Error)\n"; # open database my $db = new BerkeleyDB::Hash( -Filename => DBPATH, -Env => $env, -Flags => DB_CREATE()) or croak "Failed to open DBPATH: '$!' ($BerkeleyDB::Error)\n"; # spawn concurrent DB mutators and just wait for them to finish for (1..NUM_CHILDREN) { spawn_child($env, $db); } for (1..NUM_CHILDREN) { my $pid = wait; }

Replies are listed 'Best First'.
Re: BerkeleyDB blowing up on multiple, concurrent processes
by perrin (Chancellor) on Jun 14, 2006 at 19:32 UTC
Re: BerkeleyDB blowing up on multiple, concurrent processes
by m0r0n (Initiate) on Jun 14, 2006 at 19:58 UTC
    Figured out what was wrong. The $env and $db vars need to be child-process-local or the locking doesn't work right. Here's the updated, working script code:
    #!/usr/bin/env perl # vim:set ts=4 sw=4 ai: use strict; use warnings; use Carp; use POSIX ":sys_wait_h"; use Data::Dumper; use lib '/junk/lib'; use BerkeleyDB; use BerkeleyDB::Hash; use Time::HiRes qw(gettimeofday tv_interval); use constant RUNS => 1000000; use constant DBPATH => '/junk/test.db'; use constant NUM_CHILDREN => 10; $|++; # print backtrace info sub bt { croak "DIED ($.)\n"; } # generate random integer sub rand_num { 4278190080 + int(rand(1048576)); } # read an entry from a BDB sub read_db { my ($db, $key) = @_; my $val; $db->db_get($key, $val); return (defined($val)) ? eval($val) : {}; } # increment an entry in a BDB, inserting if not there sub write_db { my ($db, $key, $stuff) = @_; my $val = read_db($db, $key); $val->{timestamp} = time; $val->{events}++; $val->{stuff}++ if ($stuff); my $str = Dumper(\$val) or croak "failed to serialize row data: $! +\n"; return $db->db_put($key, $str); } # fork off a worker process that will mutate the BDB independently of # the others sub spawn_child { my $pid = fork; croak "Could not fork: '$!'\n" unless (defined($pid)); return unless ($pid == 0); my $env = new BerkeleyDB::Env( -Cachesize => 4 * 1024 * 1024, -Flags => (DB_CREATE() | DB_INIT_CDB() | DB_INIT_MPOOL())) or croak "Failed to create DB env: '$!' ($BerkeleyDB::Error)\n +"; my $db = new BerkeleyDB::Hash( -Filename => DBPATH, -Env => $env, -Flags => DB_CREATE()) or croak "Failed to open DBPATH: '$!' ($BerkeleyDB::Error)\n"; # child process starts here my ($start, $finish, $key, $ip, $rv, $val, $reads, $writes, $delet +es); # perform a number of operations using BDB transactions $start = [ gettimeofday ]; for (1..RUNS) { $key = int(rand(101)); $ip = rand_num(); if ($key <= 49) { # do random update/insert $rv = write_db($db, $ip, ($key % 2 == 0)); croak "db_put failed: $rv ($! / $BerkeleyDB::Error)\n" unless ($rv == 0); $reads++; $writes++; } elsif ($key <= 98) { # do random read read_db($db, $ip); $reads++; } else { # do random delete $db->db_del($key); $deletes++; } } $finish = [ gettimeofday ]; # print summary info for timing my $sum = ($reads || 0) + ($writes || 0) + ($deletes || 0); my $elapsed = tv_interval($start, $finish); printf "$$: %d reads, %d writes, %d deletes (%d total) in %0.2f se +conds: %0.2f ops/sec\n", ($reads || 0), ($writes || 0), ($deletes || 0), $sum, $elapsed, $sum / $elapsed; exit 0; } # install signal handlers $SIG{__WARN__} = \&bt; # spawn concurrent DB mutators and just wait for them to finish for (1..NUM_CHILDREN) { spawn_child(); } for (1..NUM_CHILDREN) { my $pid = wait; }
Re: BerkeleyDB blowing up on multiple, concurrent processes
by ioannis (Abbot) on Jun 14, 2006 at 19:31 UTC
    The first thing I would do is read the logs; enable logging with DB_INIT_LOG .

Log In?
Username:
Password:

What's my password?
Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: perlquestion [id://555331]
Approved by friedo
help
Chatterbox?
and the web crawler heard nothing...

How do I use this?Last hourOther CB clients
Other Users?
Others wandering the Monastery: (7)
As of 2024-04-26 08:48 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found