diff options
Diffstat (limited to 'f3s/r-nodes/Rexfile')
| -rw-r--r-- | f3s/r-nodes/Rexfile | 35 |
1 files changed, 31 insertions, 4 deletions
diff --git a/f3s/r-nodes/Rexfile b/f3s/r-nodes/Rexfile index 846b539..0dc2aea 100644 --- a/f3s/r-nodes/Rexfile +++ b/f3s/r-nodes/Rexfile @@ -10,8 +10,14 @@ use Rex -feature => [ '1.14', 'exec_autodie' ]; use Rex::Logger; use File::Basename qw(dirname); use File::Spec::Functions qw(catfile rel2abs); +use Cwd qw(realpath); -my $RNODES_DIR = dirname( rel2abs(__FILE__) ); +# Rex loads the Rexfile as a synthetic module (__Rexfile__.pm) via @INC, so +# __FILE__ resolves to the internal Rex loader path rather than this file. +# $::rexfile is set to $0 (the -f argument) in Rex::CLI before any tasks run; +# realpath() resolves any relative component against the CWD at load time so +# the path remains valid even when Rex forks worker processes for parallelism. +my $RNODES_DIR = dirname( realpath($::rexfile) ); # All three k3s Rocky Linux VMs; root SSH is configured via authorized_keys. group r_nodes => qw( @@ -26,14 +32,16 @@ sudo FALSE; # Deploy in parallel — tasks are idempotent and independent per node. parallelism 3; -# Deploy the NFS mount health-monitor script and its systemd units to -# all three r-nodes, then reload systemd and restart the timer so the -# new files take effect immediately. +# Deploy the NFS mount health-monitor script, its systemd units, and the +# tunable configuration file to all three r-nodes, then reload systemd and +# restart the timer so the new files take effect immediately. # # Files managed: # /usr/local/bin/check-nfs-mount.sh (monitor + auto-repair script) +# /etc/default/nfs-mount-monitor (tunable: NFS_FAIL_THRESHOLD) # /etc/systemd/system/nfs-mount-monitor.service # /etc/systemd/system/nfs-mount-monitor.timer +# /var/lib/nfs-mount-monitor/ (state dir for fail-count file) # # Idempotent: Rex only writes the file when content changes; the # on_change handler reloads systemd and restarts the timer only when @@ -47,6 +55,14 @@ task 'nfs_mount_monitor', # Reload flag — set to 1 if any file changed, so we only reload once. my $changed = 0; + # Ensure the state directory for the fail counter exists with tight + # permissions (only root should read/write the counter). + file '/var/lib/nfs-mount-monitor', + ensure => 'directory', + owner => 'root', + group => 'root', + mode => '700'; + # Deploy the health-monitor script. file '/usr/local/bin/check-nfs-mount.sh', source => catfile( $monitor_dir, 'check-nfs-mount.sh' ), @@ -55,6 +71,17 @@ task 'nfs_mount_monitor', mode => '755', on_change => sub { $changed = 1 }; + # Deploy the tunable configuration (NFS_FAIL_THRESHOLD). + # The leading '-' in EnvironmentFile=-/etc/default/... means systemd + # tolerates the file being absent, but we deploy it so the threshold + # is explicitly documented on each node. + file '/etc/default/nfs-mount-monitor', + source => catfile( $monitor_dir, 'nfs-mount-monitor.default' ), + owner => 'root', + group => 'root', + mode => '644', + on_change => sub { $changed = 1 }; + # Deploy the systemd service unit. file '/etc/systemd/system/nfs-mount-monitor.service', source => catfile( $monitor_dir, 'nfs-mount-monitor.service' ), |