RepoMirrors
/
mars
mirror of https://github.com/schoebel/mars
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

marsadm: workaround hanging systemd

This commit is contained in:
Thomas Schoebel-Theuer 2019-02-07 06:52:15 +01:00 committed by Thomas Schoebel-Theuer
parent 9440d2b5e8
commit 147e29049e
1 changed files with 20 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
userspace/marsadm
View File

@ -1239,11 +1239,12 @@ sub check_todo {
}
sub check_status {
my ($cmd, $res, $key, $val, $wait, $unchecked, $inv, $peer) = @_;
my ($cmd, $res, $key, $val, $wait, $unchecked, $inv, $peer, $action) = @_;
$peer = $host unless defined($peer);
my ($key_msg, $val_msg, $wait_msg, $action_msg) = _make_messages(@_);
my $path = correct_path("$mars/resource-$res/actual-$peer/$key");
my $link;
my $rounds = 0;
for (;;) {
$link = get_link($path, $unchecked);
$link = 0 unless (defined($link) && $link ne "");
@ -1257,6 +1258,12 @@ sub check_status {
ldie "Cannot execute $cmd on resource $res: actual '$key_msg' must be $val_msg. $action_msg Also ensure that your command _can_ succeed.\n" if !$wait;
}
sleep_timeout();
$rounds++;
if (defined($action) && $action && $rounds > 1) {
lprint "action: $action\n" if $verbose;
eval "$action";
$rounds = 0;
}
}
lprint "OK at $peer: '$path' has acceptable value '$link'\n";
}
@ -3399,7 +3406,18 @@ sub primary_phase0b {
my $connect_path = "$mars/resource-$res/todo-$host/connect";
my $connect = get_link($connect_path, 1);
return if (!defined($connect) || !$connect);
check_status($cmd, $res, "open-count", 0, 1, undef, undef, $old);
# Notice: this is a workaround for a problem which is
# outside of our scope. For example, a remote umount will
# fail when any (delayed) process has any filehandle open.
# As long as the umount was unsuccessful, we _cannot_ proceed
# with planned handover.
# As a workaround, we constantly trigger the remote systemd
# in the _hope_ that the umount will succeed, and the
# open-count will then go down to zero, hopefully somewhen.
my $watch = "$mars/resource-$res/systemd-want";
my $action = "";
$action = "system(\"touch -h $watch\");" if -l $watch;
check_status($cmd, $res, "open-count", 0, 1, undef, undef, $old, $action);
}
# when necessary, switch to secondary (intermediately)