fix: reintroduce NHC config template (#27)

Co-authored-by: jamesbeedy <[email protected]>

charms/slurmctld/src/constants.py

@@ -15,7 +15,7 @@
     "GresTypes": "gpu",
     "HealthCheckInterval": "600",
     "HealthCheckNodeState": "ANY,CYCLE",
-    "HealthCheckProgram": "/usr/sbin/omni-nhc-wrapper",
+    "HealthCheckProgram": "/usr/sbin/charmed-nhc-wrapper",
     "MailProg": "/usr/bin/mail.mailutils",
     "PluginDir": "/usr/lib/x86_64-linux-gnu/slurm-wlm",
     "PlugStackConfig": "/etc/slurm/plugstack.conf.d/plugstack.conf",

charms/slurmd/.wokeignore

@@ -0,0 +1 @@
+src/templates/nhc.conf.tmpl

charms/slurmd/src/slurmd_ops.py

@@ -166,6 +166,7 @@ def install(self) -> bool:
             return False
 
         self.render_nhc_config()
+        self.render_nhc_wrapper("")
 
         spool_dir = Path("/var/spool/slurmd")
         spool_dir.mkdir()
@@ -271,10 +272,10 @@ def render_nhc_config(self, extra_configs=None) -> bool:
         return True
 
     def render_nhc_wrapper(self, params: str) -> None:
-        """Render the /usr/sbin/omni-nhc-wrapper script."""
-        logger.debug(f"## rendering /usr/sbin/omni-nhc-wrapper: {params}")
+        """Render the /usr/sbin/charmed-nhc-wrapper script."""
+        logger.debug(f"## rendering /usr/sbin/charmed-nhc-wrapper: {params}")
 
-        target = Path("/usr/sbin/omni-nhc-wrapper")
+        target = Path("/usr/sbin/charmed-nhc-wrapper")
 
         target.write_text(
             textwrap.dedent(

charms/slurmd/src/templates/nhc.conf.tmpl

@@ -0,0 +1,170 @@
+# Enforce short hostnames to match the node names as tracked by slurm.
+* || HOSTNAME="$HOSTNAME_S"
+
+#######################################################################
+###
+### NHC Configuration Variables
+###
+# Explicitly instruct NHC to assume PBS (TORQUE, PBSPro) is the Resource Manager
+#   * || export NHC_RM=pbs
+
+# Do not mark nodes offline
+#   * || export MARK_OFFLINE=0
+
+# Activate debugging mode
+#   * || export DEBUG=1
+
+# Set watchdog timer to 15 seconds
+#   * || export TIMEOUT=15
+
+# In out-of-band contexts, enable all checks
+#   * || export NHC_CHECK_ALL=1
+
+# Make sure $PATH contains important directories for diagnostic commands
+#   * || export MOABHOMEDIR="/opt/moab"
+#   * || export PATH="$MOABHOMEDIR/bin:$PATH"
+
+
+#######################################################################
+###
+### Hardware checks
+###
+# Set these to your correct socket, core, and thread counts.
+#   * || check_hw_cpuinfo 2 12 24
+
+# Set these to the amount of physical RAM you have (leave the fudge factor).
+#   * || check_hw_physmem 32gb 32gb 5%
+
+# Set these to the amount of swap you have (leave the fudge factor).
+#   * || check_hw_swap 8g 8g 3%
+
+# If you prefer to use this instead of the previous two, you can.
+#   * || check_hw_mem 40g 40g 5%
+
+# Check specifically for free physical memory.
+#   * || check_hw_physmem_free 1MB
+
+# Same, but for swap space.
+#   * || check_hw_swap_free 1MB
+
+# Check for some sort of free memory of either type.
+   * || check_hw_mem_free 1mb
+
+# Checks that there's a QDR IB interface that's ACTIVE and shows LinkUp.
+#   * || check_hw_ib 40
+
+# Checks for an active Myrinet interface named "myri0."
+#   * || check_hw_gm myri0
+
+# Checks for an active ethernet interface named "eth1."
+#   * || check_hw_eth eth1
+
+# Make sure we're running the correct BIOS version on all nodes.
+#   * || check_dmi_data_match "BIOS Information: Version: 2.0.1"
+
+# Make sure our RAM is running at the correct bus rate.
+#   * || check_dmi_data_match -t "Memory Device" "*Speed: 1866 MHz"
+
+# Check the mcelog daemon for any pending errors.
+#   * || check_hw_mcelog
+
+
+#######################################################################
+###
+### Filesystem checks
+###
+# All nodes should have their root filesystem mounted read/write.
+   * || check_fs_mount_rw -f /
+
+# Assert that /tmp is a mounted filesystem of type "tmpfs."
+#   * || check_fs_mount_rw -t tmpfs -f /tmp
+
+# Controlling TTYs are a good thing!
+#   * || check_fs_mount_rw -t devpts -s '/(none|devpts)/' -f /dev/pts
+
+# Make sure the root filesystem doesn't get too full.
+#   * || check_fs_free / 3%
+
+# Free inodes are also important.
+#   * || check_fs_ifree / 1k
+
+# The following illustrates how to assert an NFSv3 mount (or any other specific mount option).
+#   * || check_fs_mount -s bluearc0:/home -t nfs -o '/(^|,)vers=3(,|$)/' -f /home
+
+
+#######################################################################
+###
+### File/metadata checks
+###
+# These should always be directories and always be read/write/execute and sticky.
+   * || check_file_test -r -w -x -d -k /tmp /var/tmp
+
+# These should always be readable and should never be empty.
+   * || check_file_test -r -s /etc/passwd /etc/group
+
+# Assert common properties for /dev/null (which occasionally gets clobbered).
+   * || check_file_test -c -r -w /dev/null /dev/zero
+#   * || check_file_stat -m 0666 -u 0 -g 0 -t 1 -T 3 /dev/null
+
+# Make sure there's relatively recent activity from the syslog.
+#   * || check_file_stat -n 7200 /var/log/messages
+
+# Validate a couple important accounts in the passwd file.
+#   * || check_file_contents /etc/passwd "/^root:x:0:0:/" "sshd:*"
+
+
+#######################################################################
+###
+### Process checks
+###
+# Everybody needs sshd running, right?  But don't use -r (restart)!
+#  * || check_ps_service -u root -S sshd
+
+# Check for munge with the correct user
+   * || check_ps_service -u munge munged
+
+# The cron daemon is another useful critter...
+#   * || check_ps_service -r crond
+
+# Check for wulfd but don't manage it.
+#   * || check_ps_daemon wulfd root
+
+# Make sure no users are SSH'd in, but don't kill them.
+#   * || check_ps_blacklist sshd '!root'
+
+# Flag and kill any processes which are owned by unauthorized users.
+#   * || check_ps_unauth_users log syslog kill
+
+# Flag any user processes not properly parented.
+#   * || check_ps_userproc_lineage log syslog
+
+# Most systems also need NFS locking services.
+#   * || check_ps_service -d rpc.statd -r nfslock
+
+# The audit daemon can sometimes disappear if things get hairy.
+#   * || check_ps_service -r auditd
+
+# This is only valid for RHEL6 and similar/newer systems.
+#   * || check_ps_service -d rsyslogd -r rsyslog
+
+# In the case of MySQL, it's typically better to cycle.
+#   * || check_ps_service -c mysqld
+
+# Double your core count is a good rule of thumb for load average max.
+#   * || check_ps_loadavg 24
+# This should work if you place it after one of the check_hw_*() checks.
+#   * || check_ps_loadavg $((2*HW_CORES))
+
+
+#######################################################################
+###
+### Other checks
+###
+# Check to verify that SELinux is disabled.  (Remove the "-r 1" to verify it's enabled.)
+#   * || check_cmd_status -t 1 -r 1 selinuxenabled
+
+# Verify settings for an Ethernet interface.
+#   * || check_cmd_output -m '/addr:10\.0\.0\.1/' -m '/Bcast:10\.0\.0\.255/' -m '/Mask:255\.255\.255\.0/' -m '/^[[:space:]]*UP /' /sbin/ifconfig eth3
+
+# nVidia HealthMon GPU health checks (requires Tesla Development Kit)
+#   * || check_nv_healthmon