you can’t just add a random tag into libvirt xml in a random place, it will be dropped by libvirt. you can add it to metadata though. we use that for ovirt-specific information

so..hmm..if IIUC you basically just need not to execute dns update(remove the entry for vm) when VM migrates away. and do execute it when it shuts down. maybe i can suggest two other approaches which could work? these would be preferable because manipulating with libvirt’s xml at the time of lifecycle changes are better to be avoided. libvirt is touching the xml at the same tie and it may run into ugly locking problems. how about - use after_vm_migrate_source and make a note of that vmid (touch a file in /tmp/<vmdid> or whatever), and then check that in after_vm_destroy or - use before_vm_destroy and use vdsm-client VM getStats vmid=‘xyz’ to get the curent VM’s status from vdsm (before it goes away) and you should see if it’s “Migration Source” vs anything else(Powering Down for ordinary shutdowns or Up for crashes, I guess) Thanks, michal

Dear Michal, We went with your first recommendation and built two python scripts. It works very nice and I am happy to share them with you: 1. after_vm_migrate_source/60_register_migration //// #!/usr/bin/python2 from __future__ import print_function import hooking import os import logging logger = logging.getLogger("register_migration") if __name__ == "__main__": try: logging.basicConfig(filename="/var/log/vdsm/custom_hooks.log", level=logging.INFO, format='%(asctime)s %(levelname)s %(name)s:%(message)s', datefmt = '%Y-%m-%d %H:%M:%S') domxml = hooking.read_domxml() vmid = domxml.getElementsByTagName('uuid')[0] vmid = vmid.firstChild.data tmp_file = os.path.join("/tmp", vmid) open(tmp_file, "a").close() logger.info("Migrating VM {vmid}, creating temp file {tmp_file}".format(vmid=vmid, tmp_file=tmp_file)) except Exception as e: logging.exception("Caught exception") //// 2. after_vm_destroy/60_nsupdate: ///// #!/usr/bin/python2 from __future__ import print_function import hooking import logging import os import subprocess from subprocess import PIPE, STDOUT logger = logging.getLogger("dns-update") if __name__ == "__main__": try: logging.basicConfig(filename="/var/log/vdsm/custom_hooks.log", level=logging.INFO, format='%(asctime)s %(levelname)s %(name)s:%(message)s', datefmt = '%Y-%m-%d %H:%M:%S') domxml = hooking.read_domxml() name = domxml.getElementsByTagName('name')[0] name = " ".join(name.nodeValue for name in name.childNodes if name.nodeType == name.TEXT_NODE) vmid = domxml.getElementsByTagName('uuid')[0] vmid = vmid.firstChild.data tmp_file = os.path.join("/tmp", vmid) if not os.path.exists(tmp_file): nsupdate_commands = """server {server_ip} update delete {vm_name}.example.com a update delete {vm_name}. example.com aaaa update delete {vm_name}. example.com txt send """.format(server_ip="1.2.3.4", vm_name=name) logger.info("Deleting VM {name} DNS records".format(name=name)) p = subprocess.Popen(["nsupdate", "-k", "/usr/libexec/vdsm/hooks/after_vm_destroy/nsupdate.private"], stdin=PIPE, stdout=PIPE, stderr=STDOUT) result = p.communicate(input=nsupdate_commands)[0] logger.info("Output:" + result) else: os.remove(tmp_file) logger.info("Ignoring DNS records delete for VM {name}:{vmid}".format(name=name, vmid=vmid)) except Exception as e: logger.exception("Caught exception") //// Both scripts are writing to same log file on each server and log lines look like this: //// 2019-11-06 13:11:12 INFO register_migration:Migrating VM 8101da66-31a1-4818-9d82-a8602d11bcf1, creating temp file /tmp/8101da66-31a1-4818-9d82-a8602d11bcf1 2019-11-06 13:11:13 INFO dns-update:Ignoring DNS records delete for VM testhooksatt2:8101da66-31a1-4818-9d82-a8602d11bcf1 2019-11-06 13:12:13 INFO dns-update:Deleting VM testhooksatt2 DNS records 2019-11-06 13:12:13 INFO dns-update:Output: //// Once again , I thank you and appreciate your help. ----- kind regards/met vrindelijke groet Marko Vrgotic ActiveVideo On 03/10/2019, 15:28, "Vrgotic, Marko" <M.Vrgotic(a)activevideo.com&gt; wrote: Michal, Thank you - that certainly helps. I will give it a try, using your suggestion and see how far I get. On 03/10/2019, 13:54, "Michal Skrivanek" <michal.skrivanek(a)redhat.com&gt; wrote: so..hmm..if IIUC you basically just need not to execute dns update(remove the entry for vm) when VM migrates away. and do execute it when it shuts down. maybe i can suggest two other approaches which could work? these would be preferable because manipulating with libvirt’s xml at the time of lifecycle changes are better to be avoided. libvirt is touching the xml at the same tie and it may run into ugly locking problems. how about - use after_vm_migrate_source and make a note of that vmid (touch a file in /tmp/<vmdid> or whatever), and then check that in after_vm_destroy or - use before_vm_destroy and use vdsm-client VM getStats vmid=‘xyz’ to get the curent VM’s status from vdsm (before it goes away) and you should see if it’s “Migration Source” vs anything else(Powering Down for ordinary shutdowns or Up for crashes, I guess) Thanks, michal