RepoMirrors/mars
1
0
Fork 0
mirror of https://github.com/schoebel/mars synced 2025-02-12 10:07:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

marsadm: fix split brain prevention by waiting on cluster

Browse Source
This commit is contained in:
Thomas Schoebel-Theuer 2014-11-29 00:42:57 +01:00
parent ffa0824e32
commit 753bcff8a8
1 changed files with 16 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
userspace/marsadm
View File

@ -209,8 +209,9 @@ unless (defined($ARGV[0]) && $ARGV[0] =~ m/cluster|cat/) {
sub get_alive_links {
my $res = shift || "all";
my $alive = shift || "alive";
my $glob = "$mars/$alive-*";
if ($res ne "all") {
my $hosts = shift || "*";
my $glob = "$mars/$alive-$hosts";
if ($res ne "all" && $hosts ne "*") {
$glob = "$mars/$alive-{";
my $count = 0;
foreach my $peer (glob("$mars/resource-$res/data-*")) {
@ -308,9 +309,9 @@ sub wait_cond {
# wait until some communication has occurred
sub wait_cluster {
return wait_cond(@_) if int(@_) >= 3;
my $cmd = shift;
my $res = shift || "all";
my $hosts = shift || "*";
my $start_time = mars_time();
_trigger();
my $delta = $timeout > 0 ? $timeout : 30;
@ -318,7 +319,7 @@ sub wait_cluster {
my $dead_count = 0;
my $alive_count = 0;
my $unknown_count = 0;
my %status = get_alive_links($res, "time");
my %status = get_alive_links($res, "time", $hosts);
my $now = mars_time();
foreach my $peer (keys(%status)) {
next if $peer eq $host;
@ -907,7 +908,9 @@ sub log_purge_res {
}
sub try_to_avoid_splitbrain {
my ($cmd, $res) = @_;
my ($cmd, $res, $old_primary) = @_;
$old_primary = "" if $old_primary eq "(none)";
wait_cluster($cmd, $res, $old_primary);
if (!detect_splitbrain($res, 0)) {
lwarn "ATTENTION: you are starting a non-forced primary switchover in a split brain situation.\n";
lwarn "ATTENTION: that's no good idea.\n";
@ -916,11 +919,11 @@ sub try_to_avoid_splitbrain {
return;
}
# now try to prevent producing a _new_ split brain situation....
my ($min, $max) = get_minmax_versions($res);
my @host_list = glob("$mars/resource-$res/replay-*");
return if scalar(@host_list) < 2;
my $vers_glob = sprintf("$mars/resource-$res/version-%09d-*", $max);
for (;;) {
my ($min, $max) = get_minmax_versions($res);
my $vers_glob = sprintf("$mars/resource-$res/version-%09d-*", $max);
my $ok = 1;
my $replay_err_path = "$mars/resource-$res/actual-$host/msg-err-replay-stop";
my $replay_err = get_link($replay_err_path, 1);
@ -1931,8 +1934,8 @@ sub primary_phase1 {
my $old = _get_designated_primary($res);
return if ($old eq $host and $cmd eq "primary");
my $new = "(none)";
try_to_avoid_splitbrain($cmd, $res, $old) if (!$force and $cmd eq "primary");
return if $old eq $new;
try_to_avoid_splitbrain(@_) if (!$force and $cmd eq "primary");
_primary_res($res, $new, $old);
}
@ -2929,13 +2932,14 @@ sub eval_fn {
wait_cond($$env{"cmd"}, $$env{"res"}, $specific);
return "";
}
if (/^wait$/) {
if (/^wait(?:[-_]?resource)?$/) {
my $specific = parse_macro($arg1, $env);
wait_cond($$env{"cmd"}, $$env{"res"}, $specific);
return "";
}
if (/^wait[-_]?resource$/) {
wait_cluster($$env{"cmd"}, $$env{"res"});
if (/^wait[-_]?cluster$/) {
my $specific = parse_macro($arg1, $env);
wait_cluster($$env{"cmd"}, $$env{"res"}, $specific);
return "";
}
# generic flow control and loops
@ -3929,7 +3933,7 @@ my %cmd_table =
],
"wait-resource"
=> [
\&wait_cluster,
\&wait_cond,
],
# compatible keywords (or their derivatives)