mirror of
https://github.com/schoebel/mars
synced 2025-03-11 07:47:41 +00:00
marsadm: add cron --autoclean
This commit is contained in:
Thomas Schoebel-Theuer
Thomas Schoebel-Theuer
parent
bc63996832
commit
f0f88b7258
@ -234,6 +234,7 @@ my $backup_dir = "$mars/backups-" . time();
|
||||
my $force = 0;
|
||||
my $ignore_sync = 0;
|
||||
my $cron_mode = 0;
|
||||
my $cron_autoclean_days = 0;
|
||||
my $timeout = 600;
|
||||
my $phase_nr = 0;
|
||||
my $ssh_port = 22;
|
||||
@ -5887,6 +5888,187 @@ sub cron_phase2 {
|
||||
return 0;
|
||||
}
|
||||
|
||||
my %protected_peers;
|
||||
|
||||
sub get_protected_peers {
|
||||
my $ips_dir = "$mars/ips";
|
||||
if (!scalar(keys(%protected_peers))) {
|
||||
foreach my $path (raw_glob("$ips_dir/ip-*")) {
|
||||
$path =~ m:/ip-(.*):;
|
||||
my $peer = $1;
|
||||
# do not protect any ill-formed peer names, by definition they need to be repaired.
|
||||
if (!$peer || $peer =~ $match_reserved_id) {
|
||||
lwarn "skipping invalid / illformed pathname '$path'\n";
|
||||
next;
|
||||
}
|
||||
$protected_peers{$peer} = 1;
|
||||
}
|
||||
if (!scalar(keys(%protected_peers))) {
|
||||
lwarn "VERY DANGEROUS: the directory '$ips_dir' is EMPTY!!!";
|
||||
}
|
||||
foreach my $peer (keys(%protected_peers)) {
|
||||
lprint "REAL PEER '$peer'\n";
|
||||
}
|
||||
}
|
||||
if (!scalar(keys(%protected_peers))) {
|
||||
ldie "Cannot determine protected hostnames / peernames in $ips_dir - ABORTING for safety!";
|
||||
} else {
|
||||
# ADDITIONAL SAFETY: don't forget any further information sources.
|
||||
if (!$protected_peers{$real_host}) {
|
||||
lwarn "Hopefully, the _reported_ real hostname '$real_host' (as reported by 'uname -a', NOT to be confused with a virtual hostname on the commandline via --host=\$other_name parameter) is REALLY correct.\n";
|
||||
lwarn "DOUBLE-CHECK or TRIPLE-CHECK whether your configured hostname as reported by 'uname -a' is REALLY CORRECT.\n";
|
||||
lwarn "Like anywhere else (not limited to MARS), ANY ERROR in the hostname configuration can easily create a DISASTER, up to DATA LOSS.\n";
|
||||
ldie "POTENTIAL DISASTER: reported real host '$real_host' is missing in $ips_dir - ABORTING FOR SAFETY\n";
|
||||
}
|
||||
if (!$protected_peers{$host}) {
|
||||
lwarn "DANGEROUS: host '$host' was missing in $ips_dir - CHECK BY HAND whether this is correct.\n";
|
||||
$protected_peers{$host} = 1;
|
||||
lprint "ADD PEER '$host'\n";
|
||||
}
|
||||
}
|
||||
# Finally, add _potentially_ _relevant_ peernames (not to be confused with _reported_ peernames).
|
||||
my @relevant_resources = get_member_resources($real_host);
|
||||
foreach my $relevant_res (@relevant_resources) {
|
||||
foreach my $path (raw_glob("$mars/resource-$relevant_res/replay-*")) {
|
||||
$path =~ m:/replay-(.*):;
|
||||
my $peer = $1;
|
||||
unless ($protected_peers{$peer}) {
|
||||
lwarn "PLEASE CHECK whether peer '$peer' or resource '$relevant_res' is SOMEHOW misconfigured.\n";
|
||||
# CHECK: should this be added?
|
||||
# Some people might forget stray replaylinks in alive resources.
|
||||
# If I add this, some stray data might be kept forever, and sysadmins might be mis-informed
|
||||
# about its severity.
|
||||
# By _not_ activating this at the moment, the sysadmins will be responsible for checking
|
||||
# where the real problem is.
|
||||
# I cannot know this, for example when stone-aged hosts have been physically decommissioned forever.
|
||||
# OTOH the --autoclean option is expected to clean up anything which is "stray" and may be
|
||||
# "irritating".
|
||||
if (0) {
|
||||
$protected_peers{$peer} = 1;
|
||||
lprint "ADD PEER '$peer'\n";
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
sub _autoclean_dir {
|
||||
my ($dir, $limit_stamp, $clean_full, $level) = @_;
|
||||
return if $level <= 0;
|
||||
get_protected_peers();
|
||||
foreach my $path (glob("$dir/*")) {
|
||||
# Some _global_ non-host-specific elements need to be protected _always_.
|
||||
# These can only deleted by filesystem destruction.
|
||||
next if $path =~ m:/(uuid|userspace|defaults|todo-global|ips)$:;
|
||||
if (!$clean_full) {
|
||||
# Keep protected elements.
|
||||
# Some non-host-specific elements need to be protected _always_.
|
||||
next if $path =~ m:/(primary|size|)$:;
|
||||
next if $path =~ m:/(log-|systemd):;
|
||||
# host-specific protections.
|
||||
my $found = 0;
|
||||
foreach my $peer (keys(%protected_peers)) {
|
||||
$found++ if $path =~ m:-$peer\Z:;
|
||||
}
|
||||
next if $found;
|
||||
}
|
||||
my $stamp = get_link_stamp($path);
|
||||
if (-d $path) {
|
||||
_autoclean_dir($path, $limit_stamp, $clean_full, $level - 1);
|
||||
} elsif (-l $path) {
|
||||
if ($stamp > 0 && $stamp < $limit_stamp) {
|
||||
lprint "AUTOCLEANING '$path'\n";
|
||||
unlink($path) unless $dry_run;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
sub autoclean_res {
|
||||
my ($cmd, $res, $after_days) = @_;
|
||||
if ($after_days <= 0) {
|
||||
lwarn "cannot $cmd: age $after_days days is lower than 1 day\n";
|
||||
return;
|
||||
}
|
||||
my $resdir = "$mars/resource-$res";
|
||||
return unless -d $resdir;
|
||||
lprint "autocleaning $res remains after $after_days days...\n" if $verbose;
|
||||
my $start_time = mars_time();
|
||||
my $limit_stamp = $start_time - $after_days * 3600 * 24;
|
||||
get_protected_peers();
|
||||
_autoclean_dir($resdir, $limit_stamp, 0, 3);
|
||||
lprint "autoclean $res done.\n" if $verbose;
|
||||
}
|
||||
|
||||
sub cron_phase3 {
|
||||
my ($cmd, $res) = @_;
|
||||
return 0 unless $cron_autoclean_days > 0;
|
||||
autoclean_res($cmd, $res, $cron_autoclean_days);
|
||||
return 0;
|
||||
}
|
||||
|
||||
sub autoclean_any {
|
||||
my ($cmd, $res, $after_days) = @_;
|
||||
if ($after_days <= 0) {
|
||||
lwarn "cannot $cmd: age $after_days days is lower than 1 day\n";
|
||||
return;
|
||||
}
|
||||
# hard coded: at least 1 month ;)
|
||||
my $min_after_days = 30;
|
||||
$after_days = $min_after_days if $after_days = $min_after_days;
|
||||
my $start_time = mars_time();
|
||||
my $limit_stamp = $start_time - $after_days * 3600 * 24;
|
||||
get_protected_peers();
|
||||
foreach my $resdir (raw_glob("$mars/resource-*")) {
|
||||
my $protect_this = 0;
|
||||
my $protected_peer_list = "";
|
||||
foreach my $peer (keys(%protected_peers)) {
|
||||
my $is_protected = scalar(raw_glob("$resdir/*-$peer"));
|
||||
if ($is_protected) {
|
||||
$protect_this = 1;
|
||||
$protected_peer_list .= "," if $protected_peer_list;
|
||||
$protected_peer_list .= $peer;
|
||||
}
|
||||
}
|
||||
if ($protect_this) {
|
||||
lprint "SKIPPING '$resdir' due to peers '$protected_peer_list'\n" if $verbose;
|
||||
next;
|
||||
}
|
||||
# Also check that everything is stone-aged ;)
|
||||
my $newest_stamp = 0;
|
||||
my $newest_element = "";
|
||||
foreach my $path (raw_glob("$resdir/*")) {
|
||||
my $age = get_stamp($path);
|
||||
if ($age >= $limit_stamp) {
|
||||
$protect_this = 1;
|
||||
if ($age > $newest_stamp) {
|
||||
$newest_stamp = $age;
|
||||
$newest_element = $path;
|
||||
}
|
||||
}
|
||||
}
|
||||
if ($protect_this) {
|
||||
lprint "SKIPPING '$resdir' due to newest element '$newest_element' having age '$newest_stamp' (since the UNIX epoch)\n" if $verbose;
|
||||
next;
|
||||
}
|
||||
if ($force && !$dry_run) {
|
||||
lprint "FULLY AUTOCLEANING $resdir remains after $after_days days...\n";
|
||||
system("rm -rf $resdir") unless $dry_run;
|
||||
} else {
|
||||
lprint "WOULD autoclean $resdir FULLY via rm -rf after $after_days days\n" if $verbose;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
sub cron_phase4 {
|
||||
my ($cmd, $res) = @_;
|
||||
return 0 unless $cron_autoclean_days > 0;
|
||||
lprint "======== EXTRA PHASE: GLOBAL AUTOCLEAN\n";
|
||||
autoclean_any($cmd, "*", $cron_autoclean_days);
|
||||
lprint "GLOBAL AUTOCLEAN finished.\n";
|
||||
return 0;
|
||||
}
|
||||
|
||||
sub attach_res_phase0 {
|
||||
my ($cmd, $res) = @_;
|
||||
return 0 if $force;
|
||||
@ -9051,6 +9233,14 @@ my %cmd_table =
|
||||
"SLEEP",
|
||||
\&cron_phase2,
|
||||
"delete old logfiles",
|
||||
\&cron_phase3,
|
||||
"autoclean gone peers (when enabled via --autoclean)",
|
||||
"CHANGE_RESOURCES",
|
||||
"ANY",
|
||||
[
|
||||
\&cron_phase4,
|
||||
"autoclean gone resources (when enabled via --autoclean)",
|
||||
],
|
||||
],
|
||||
"log-purge-all"
|
||||
=> [
|
||||
@ -9903,6 +10093,9 @@ foreach my $arg (@ARGV) {
|
||||
} elsif ($arg =~ s/--logger\s*=\s*(.*)/$1/) {
|
||||
$logger = $arg;
|
||||
next;
|
||||
} elsif ($arg =~ s/--autoclean\s*=\s*([0-9]+)/$1/) {
|
||||
$cron_autoclean_days = $arg;
|
||||
next;
|
||||
} elsif ($arg =~ s/--timeout\s*=\s*([0-9]+)/$1/) {
|
||||
$timeout = $arg;
|
||||
next;
|
||||
@ -10346,6 +10539,10 @@ if (ref($func) eq "ARRAY") {
|
||||
} elsif ($memb_func eq "SLEEP") {
|
||||
$memb_func = shift @list;
|
||||
sleep(7);
|
||||
} elsif ($memb_func eq "CHANGE_RESOURCES") {
|
||||
$res = shift @list;
|
||||
next;
|
||||
#$memb_func = shift @list;
|
||||
}
|
||||
}
|
||||
# nested arrays may be used for _global_ workers
|
||||
|
||||
Loading…
Reference in New Issue
Block a user