WIP install 8/n: run installer, go see inside and detect success

FIXME:
- specification of answerfile contents and dom0 cmdline is tightly linked,
  e.g. we rely on atexit=shell
- repeatedly polling with grep is junk

Author: ydirson

conftest.py

@@ -268,20 +268,33 @@ def iso_remaster():
     SOURCE_ISO = "/home/user/iso/xcp-ng-8.2.1-20231130.iso" # FIXME dict in data.py
     ANSWERFILE_URL = "http://pxe/configs/custom/ydi/install-8.2-uefi-iso-ext.xml" # FIXME
 
-    from data import ISOSR_SRV, ISOSR_PATH, TOOLS
+    from data import ISOSR_SRV, ISOSR_PATH, TEST_SSH_PUBKEY, TOOLS
     assert "iso-remaster" in TOOLS
     iso_remaster = TOOLS["iso-remaster"]
     assert os.access(iso_remaster, os.X_OK)
 
     with tempfile.TemporaryDirectory() as isotmp:
         remastered_iso = os.path.join(isotmp, "image.iso")
         iso_patcher_script = os.path.join(isotmp, "iso-patcher")
+        img_patcher_script = os.path.join(isotmp, "img-patcher")
 
         logging.info("Remastering %s to %s", SOURCE_ISO, remastered_iso)
 
+        # generate install.img-patcher script
+        with open(img_patcher_script, "xt") as patcher_fd:
+            # FIXME insert nswerfile in there
+            print(f"""#!/bin/bash
+set -ex
+INSTALLIMG="$1"
+mkdir -p "$INSTALLIMG/root/.ssh"
+echo "{TEST_SSH_PUBKEY}" > "$INSTALLIMG/root/.ssh/authorized_keys"
+""",
+                  file=patcher_fd)
+            os.chmod(patcher_fd.fileno(), 0o755)
+
         # generate iso-patcher script
         with open(iso_patcher_script, "xt") as patcher_fd:
-            passwd = "passw0rd" # FIXME hash
+            passwd = "passw0rd" # FIXME use invalid hash
             print(f"""#!/bin/bash
 set -ex
 ISODIR="$1"
@@ -297,6 +310,7 @@ def iso_remaster():
 
         # do remaster
         local_cmd([iso_remaster,
+                   "--install-patcher", img_patcher_script,
                    "--iso-patcher", iso_patcher_script,
                    SOURCE_ISO, remastered_iso
                    ])

data.py-dist

@@ -6,6 +6,9 @@
 HOST_DEFAULT_USER = "root"
 HOST_DEFAULT_PASSWORD = ""
 
+# Public key for a private key available to the test runner
+TEST_SSH_PUBKEY = "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDKz9uQOoxq6Q0SQ0XTzQHhDolvuo/7EyrDZsYQbRELhcPJG8MT/o5u3HyJFhIP2+HqBSXXgmqRPJUkwz9wUwb2sUwf44qZm/pyPUWOoxyVtrDXzokU/uiaNKUMhbnfaXMz6Ogovtjua63qld2+ZRXnIgrVtYKtYBeu/qKGVSnf4FTOUKl1w3uKkr59IUwwAO8ay3wVnxXIHI/iJgq6JBgQNHbn3C/SpYU++nqL9G7dMyqGD36QPFuqH/cayL8TjNZ67TgAzsPX8OvmRSqjrv3KFbeSlpS/R4enHkSemhgfc8Z2f49tE7qxWZ6x4Uyp5E6ur37FsRf/tEtKIUJGMRXN XCP-ng CI"
+
 # The following prefix will be added to the `name-label` parameter of XAPI objects
 # that the tests will create or import, such as VMs and SRs.
 # Default value: [your login/user]

tests/install/test_install.py

@@ -1,5 +1,12 @@
 import logging
+import os
 import pytest
+import time
+
+from lib import commands, pxe
+from lib.common import wait_for
+from lib.host import Host
+from lib.pool import Pool
 
 @pytest.mark.vm_definitions(
     dict(name="vm 1",
@@ -21,5 +28,136 @@ def test_install_nested_821_uefi(self, create_vms, iso_remaster):
         assert len(create_vms) == 1
         host_vm = create_vms[0]
 
+        vif = host_vm.vifs()[0]
+        mac_address = vif.param_get('MAC')
+        logging.info("Host VM has MAC %s", mac_address)
+
         host_vm.create_cd_vbd("xvdd")
         host_vm.insert_cd(iso_remaster)
+
+        host_vm.start()
+        wait_for(host_vm.is_running, "Wait for host VM running")
+        try:
+            # catch host-vm IP address
+            wait_for(lambda: pxe.arp_addresses_for(mac_address),
+                     "Wait for DHCP server to see Host VM in ARP tables",
+                     timeout_secs=10*60)
+            ips = pxe.arp_addresses_for(mac_address)
+            logging.info("Host VM has IPs %s", ips)
+            assert len(ips) == 1
+            host_vm.ip = ips[0]
+
+            host_vm.ssh(["ls"])
+            logging.info("ssh works")
+
+            # wait for "yum install" phase to finish
+            wait_for(lambda: host_vm.ssh(["grep",
+                                          "'DISPATCH: NEW PHASE: Completing installation'",
+                                          "/tmp/install-log"],
+                                         check=False, simple_output=False,
+                                         ).returncode == 0,
+                     "Wait for rpm installation to succeed",
+                     timeout_secs=40*60) # FIXME too big
+
+            # wait for install to finish
+            wait_for(lambda: host_vm.ssh(["grep",
+                                          "'The installation completed successfully'",
+                                          "/tmp/install-log"],
+                                         check=False, simple_output=False,
+                                         ).returncode == 0,
+                     "Wait for installation to succeed",
+                     timeout_secs=40*60) # FIXME too big
+
+            # powercycle, catch any change of IP
+            logging.info("Rebooting Host VM after successful installation")
+            try:
+                # use "poweroff" because "reboot" would cause ARP and
+                # SSH to be checked before host is down, and require
+                # ssh retries
+                host_vm.ssh(["poweroff"])
+            except commands.SSHCommandFailed as e:
+                # ignore connection closed by reboot
+                if e.returncode == 255 and "closed by remote host" in e.stdout:
+                    logging.info("sshd closed the connection")
+                    pass
+                else:
+                    raise
+            wait_for(host_vm.is_halted, "Wait for host VM halted")
+            host_vm.eject_cd()
+
+            # FIXME: make a snapshot here
+
+            # FIXME: evict MAC from ARP cache first?
+            host_vm.start()
+            wait_for(host_vm.is_running, "Wait for host VM running")
+
+            ips = pxe.arp_addresses_for(mac_address)
+            logging.info("Host VM has IPs %s", ips)
+            assert len(ips) == 1
+            host_vm.ip = ips[0]
+
+            wait_for(lambda: not os.system(f"nc -zw5 {host_vm.ip} 22"),
+                     "Wait for ssh back up on Host VM", retry_delay_secs=5)
+
+            # pool master must be reachable here
+            # FIXME: not sure why we seem to need this, while port 22 has been seen open
+            tries = 5
+            while True:
+                try:
+                    pool = Pool(host_vm.ip)
+                except commands.SSHCommandFailed as e:
+                    if "Connection refused" not in e.stdout:
+                        raise
+                    tries -= 1
+                    if tries:
+                        logging.warning("retrying connection to pool master")
+                        time.sleep(2)
+                        continue
+                    # retries failed
+                    raise
+                # it worked!
+                break
+
+            # wait for XAPI
+            # FIXME: flaky, must check logs extraction on failure
+            for service in ["control-domain-params-init",
+                            "network-init",
+                            "storage-init",
+                            "generate-iscsi-iqn",
+                            "create-guest-templates",
+                            ]:
+                try:
+                    wait_for(lambda: pool.master.ssh(["test", "-e", f"/var/lib/misc/ran-{service}"],
+                                                     check=False, simple_output=False,
+                                                     ).returncode == 0,
+                             f"Wait for ran-{service} stamp")
+                except TimeoutError:
+                    logging.warning("investigating lack of ran-{service} stamp")
+                    out = pool.master.ssh(["systemctl", "status", service], check=False)
+                    logging.warning("service status: %s", out)
+                    out = pool.master.ssh(["grep", "-r", service, "/var/log"], check=False)
+                    logging.warning("in logs: %s", out)
+
+            wait_for(pool.master.is_enabled, "Wait for XAPI to be ready", timeout_secs=30 * 60)
+
+            logging.info("Powering off pool master")
+            try:
+                # use "poweroff" because "reboot" would cause ARP and
+                # SSH to be checked before host is down, and require
+                # ssh retries
+                pool.master.ssh(["poweroff"])
+            except commands.SSHCommandFailed as e:
+                # ignore connection closed by reboot
+                if e.returncode == 255 and "closed by remote host" in e.stdout:
+                    logging.info("sshd closed the connection")
+                    pass
+                else:
+                    raise
+
+            wait_for(host_vm.is_halted, "Wait for host VM halted")
+
+        except Exception as e:
+            logging.critical("caught exception %s", e)
+            #wait_for(lambda: False, 'Wait "forever"', timeout_secs=100*60)
+            host_vm.shutdown(force=True)
+            raise