machines: Add LibvirtDBus provider #9240

KKoukiou · 2018-05-25T13:28:20Z

This patch series is rewriting martinpitt@329e623 and implements a second provider for Libvirt using Libvirt's D-Bus API (https://github.com/libvirt/libvirt-dbus/).
Since there is a lot of code that can be reused from libvirt.es6 file in the new provider, all reusable functions where moved to libvirt-common.es6 file to avoid code duplicity.

Checklist:

~~use DBus in ovirt/provider.es6 if available (so same story as for pkg/machines)~~ (postponed to follow-up)
integration tests (refactor recent check-machines and check-ovirt tests to run ones for each of the providers)
user documentation to install libvirt-dbus
farther refactor to share as much as possible between libvirt.es6 and libvirt-dbus.es6
adjust dependencies to libvirt-dbus | libvirt-client ?
fix testAddDisk (machines: Robustify testAddDisk #9952 )
get fixed libvirt-dbus on fedora-28 (Image refresh for fedora-28 #9972)
fix temporary added disks to actually go away on shutdown

pkg/machines/index.es6

mareklibra

Thank you for this effort, moving away from virsh is definately good way in the long-term!
Huray for the just released libvirt-dbus package.

Please correct me, but libvirt-dbus has still long way to pass to become a standard part of the supported distributions (aka debian or RHEL).
It's ok but just to be clear here: we will need to support both virsh and dbus connections in parallel for non-trivial time (even years?).

Adding Checklist to the PR description.

mareklibra · 2018-05-28T09:19:55Z

pkg/machines/index.es6

+                    setVirtProvider(LibvirtDbus);
+                } else {
+                    logDebug("index.es6: Setting virsh as virt provider.");
+                    setVirtProvider(Libvirt);


ovirt/provider.es6 will need to use DBus as well, if available.

(ovirt/index.es6, resp. inheritance in ovirt/provider.es6:40)

mareklibra · 2018-05-28T09:22:53Z

pkg/machines/index.es6

+                "ListActivatableNames")
+            .done(services => {
+                if (services[0].includes("org.libvirt")) {
+                    logDebug("index.es6: Setting libvirt-dbus as virt provider.");


I would use console.info() for this important decision

mareklibra · 2018-05-28T09:23:01Z

pkg/machines/index.es6

+                    logDebug("index.es6: Setting libvirt-dbus as virt provider.");
+                    setVirtProvider(LibvirtDbus);
+                } else {
+                    logDebug("index.es6: Setting virsh as virt provider.");


pkg/machines/index.es6

mareklibra · 2018-05-28T09:54:20Z

pkg/machines/libvirt-dbus.es6

+        return true;
+    },
+
+    canReset: (vmState) => vmState == 'running' || vmState == 'idle' || vmState == 'paused',


Can following be shared?

mareklibra · 2018-05-28T09:55:10Z

pkg/machines/libvirt-dbus.es6

+    serialConsoleCommand: ({
+        vm
+    }) => vm.displays['pty'] ? [ 'virsh', ...VMS_CONFIG.Virsh.connections[vm.connectionName].params, 'console', vm.name ] : false,
+


... Up to here

mareklibra · 2018-05-28T10:22:15Z

pkg/machines/libvirt-dbus.es6

+const _ = cockpit.gettext;
+
+let clientLibvirt = {
+    'session': null,


No need to list them here. They will be undefined by default, so same semantics. Their name might vary based on the configuration.

KKoukiou · 2018-06-14T12:31:31Z

@mareklibra @martinpitt @Antique I reposted the patches with some fixes, adjusted the tests to run with both providers and some added info about libvirt-dbus in plugin docs.
Can you please take a look?

The tests are all passing locally, though with the current images they will fail here, because the available libvirt-dbus release is missing some APIs that I am using in these commits. (See more here https://www.redhat.com/archives/libvir-list/2018-June/msg00898.html)

So, although we shouldn't merge this patch series without the new release of libvirt-dbus to be included in the test images, can you please review so that I can get it to a state ready to merge?

Regarding supporting the using the libvirt-dbus in oVirt package, I suggest to keep it in separate PR, since current is already very big. WDYT?

KKoukiou · 2018-06-29T17:00:29Z

Changes from previous push:

Rebased on top on mlibra's PR for Adding disks, https://github.com/machines: Create and Attach disk to a virtual machine #9166
Added support for VCPUS

KKoukiou · 2018-06-29T17:25:29Z

pkg/machines/libvirt-dbus.es6

+    },
+
+    SET_VCPU_SETTINGS ({ name, connectionName, count, max, sockets, cores, threads, isRunning }) {
+        logDebug(`${this.name}.SET_VCPU_SETTINGS(${name}):`);


I forgot to implement this :|
I 'll repost.

martinpitt · 2018-07-20T06:08:17Z

#9692 landed, can you please rebase?

KKoukiou · 2018-07-20T13:27:42Z

Ready for review. Only testAddDisk for libvirt-dbus provider is failing but I am not sure why at this point.

I am afraid that refreshing storagePools here https://github.com/cockpit-project/cockpit/blob/master/pkg/machines/components/diskAdd.jsx#L425 is not actually happening, since the storagePools object it not fetched from this props every time it's updated. (Though I am total beginner with react, feedback would be appreciated)

phrdina

Overall looking good but there are some minor issues that needs to be addressed.

phrdina · 2018-07-27T08:03:33Z

pkg/machines/index.es6

+                let providerVal = libvirtDBusavailable ? LibvirtDbus : Libvirt;
+
+                client.close();
+                return providerVal;


Just a small nit, this can be simplified by removing providerVal and using this return:
return libvirtDBusavailable ? LibvirtDbus : Libvirt;

And you can change libvirtDBusavailable into libvirtDBusAvailable.

phrdina · 2018-07-27T08:49:05Z

pkg/machines/libvirt-dbus.es6

+        return dispatch => {
+            clientLibvirt[connectionName].call(objPath, 'org.libvirt.Domain', 'GetXMLDesc', [0], TIMEOUT)
+                    .done(domXml => {
+                        let updatedXml = updateNetworkIfaceState(domXml[0], networkMac, state);


updateNetworkIfaceState can return nothing if parsin domXml fails, we should check that

phrdina · 2018-07-27T09:15:48Z

pkg/machines/libvirt-dbus.es6

+        connectionName
+    }) {
+        return dispatch => {
+            clientLibvirt[connectionName].call(objPath, 'org.libvirt.Domain', 'GetXMLDesc', [0], TIMEOUT)


GetXMLDesc can fail, shouldn't we check that?

Where is it checked? I don't see a .catch handler here?

pkg/machines/libvirt-dbus.es6

phrdina · 2018-07-27T10:15:57Z

pkg/machines/libvirt-dbus.es6

+                                clientLibvirt[connectionName].call(storageVolPath[0], 'org.libvirt.StorageVol', 'Delete', [0], TIMEOUT)
+                                        .finally(undefine_synced.bind(null, i + 1));
+                            })
+                else /* VIR_DOMAIN_UNDEFINE_MANAGED_SAVE */


We probably should use VIR_DOMAIN_UNDEFINE_SNAPSHOTS_METADATA and VIR_DOMAIN_UNDEFINE_NVRAM flags as well, otherwise cockpit would not be able to undefine such domains.

phrdina · 2018-07-27T11:11:02Z

test/verify/check-machines-dbus

@@ -0,0 +1,67 @@
+#!/usr/bin/python2


I would rather use python3 for new code, every other verify script uses python3.

phrdina · 2018-07-27T11:17:26Z

test/verify/check-machines-dbus

+
+    def detectLibvirtDbus(self):
+        m = self.machine
+        namesList = m.execute("dbus-send --system --dest=org.freedesktop.DBus "


Cannot we use python dbus module?

Not really, this code is running on the machine you execute test tests from, m.execute executes the commands inside the VM the tests are going to run inside. If I was using dbus python module I would need to run it from a script that was running inside machine that contains the dbus I want to poll.. Or correct me if I am wrong.

Oh right, I missed the fact the it is executed on a different machine.

phrdina · 2018-07-27T11:19:52Z

test/verify/machineslib.py

@@ -528,6 +524,8 @@ def testAddDisk(self):
        b.wait_present("#vm-subVmTest1-disks") # wait for the tab
        b.click("#vm-subVmTest1-disks") # open the "Disks" subtab

+        time.sleep(2);


Why we need this change? There is nothing in commit message about it.

I removed this already, I was debugging something.

phrdina · 2018-07-27T11:20:56Z

doc/guide/feature-virtualmachines.xml

-        access to Libvirtd is wrapped by the <ulink url="https://libvirt.org/virshcmdref.html">virsh</ulink> tool.</para>
+        access to Libvirtd is wrapped either by the <ulink url="https://libvirt.org/virshcmdref.html">virsh</ulink> tool or
+        <ulink url="https://libvirt.org/dbus.html">libvirt D-Bus API bindings</ulink>, depending if the latter is installed on
+        the system. If libvirt-dbus package is not available in your distrubution's software repositories you can follow the instructions


s/distrubution/distribution/

phrdina · 2018-07-27T11:22:43Z

doc/guide/feature-virtualmachines.xml

@@ -21,6 +25,15 @@
    </para>
    </section>

+    <section id="feature-virtualmachines-sessionaccess">
+    <title>Managing VMs on session connection</title>


Session access is allowed be default, I guess that you mean system access/connection.

The libvirt-dbus API has `StorageVolLookupByPath` but not by target. Populate storage list with paths instead of targets to allow easy usage of the API. Current usage with `virsh undefine` is still going to work with storage paths passed to the `--storage` parameter.

A new provider is going to be introduced in followup patch and we need to clarify from the file name the provider used in each case.

The id parameter is going to be used by followup changes introducing new provider.

Replace parseXML from jquery with DOMParser.

KKoukiou · 2018-09-17T19:01:56Z

@KKoukiou : I just landed #9945, and now this conflicts. Can you please rebase?

Done

martinpitt · 2018-09-17T20:25:54Z

Related test failures:

testAddDisk on debian-testing and fedora-29
testLibvirt on fedora-29
testCreate on fedora-i386

There are also some failures in the virsh tests, but these are (presumably) not new. But we shouldn't introduce more flakes, it's hard enough to get the tests to green as it is..

KKoukiou · 2018-09-17T22:16:02Z

Related test failures:

* testAddDisk on [debian-testing](https://209.132.184.41:8493/logs/pull-9240-20180917-191010-eb76890f-verify-debian-testing/log.html#104) and [fedora-29](https://209.132.184.41:8493/logs/pull-9240-20180917-191212-eb76890f-verify-fedora-29/log.html#104)

Looks like race condition when tests are executed very fast with state not being up to date.. I added some logs in react code and I can't reproduce locally any more.

* testLibvirt on [fedora-29](https://209.132.184.41:8493/logs/pull-9240-20180917-191212-eb76890f-verify-fedora-29/log.html#111)

Not reproducible locally. I tried 10 times with no error :/

* testCreate on [fedora-i386](https://209.132.184.41:8493/logs/pull-9240-20180917-190909-eb76890f-verify-fedora-i386/log.html#49)

This test can't be affected by this PR because the code for CREATE action is shared 1000% between the two providers.
I tried to clean a bit however the test steps and let's see it it can reappear.

There are also some failures in the virsh tests, but these are (presumably) not new. But we shouldn't introduce more flakes, it's hard enough to get the tests to green as it is..

martinpitt · 2018-09-18T07:39:55Z

Hmkay, retrying the tests a few times.

By the way, if you want to add libvirt-dbus to the rhel-x image, now would be a good time for it. I just spent my entire morning yesterday to fix the current regressions, so we got a fresh build yesterday. There's a very high chance that a build today will also succeed :-)

martinpitt · 2018-09-18T11:52:01Z

I can locally reproduce this failure on fedora-29 in about one out of three (parallel) runs. Just keeping the log pointer here for now.

KKoukiou · 2018-09-18T11:58:39Z

Reposting with the following diff. (forgot to change some naughy files)

[kkoukiou@localhost cockpit]$ git diff
diff --git a/bots/naughty/rhel-7/7891-selinux-denies-virtlogd-dbus-2 b/bots/naughty/rhel-7/7891-selinux-denies-virtlogd-dbus-2
index cf5808fb7..292a9aa0f 100644
--- a/bots/naughty/rhel-7/7891-selinux-denies-virtlogd-dbus-2
+++ b/bots/naughty/rhel-7/7891-selinux-denies-virtlogd-dbus-2
@@ -1,5 +1,5 @@
 Traceback (most recent call last):
-  File "test/verify/check-machines", line *, in testBasic
+  File "test/verify/machineslib.py", line *, in testBasic
     wait(lambda: "Linux version" in self.machine.execute("cat {0}".format(args["logfile"])), delay=3)
 *
 Error: Condition did not become true.
diff --git a/bots/naughty/rhel-x/7891-selinux-denies-virtlogd-dbus-2 b/bots/naughty/rhel-x/7891-selinux-denies-virtlogd-dbus-2
index cf5808fb7..292a9aa0f 100644
--- a/bots/naughty/rhel-x/7891-selinux-denies-virtlogd-dbus-2
+++ b/bots/naughty/rhel-x/7891-selinux-denies-virtlogd-dbus-2
@@ -1,5 +1,5 @@
 Traceback (most recent call last):
-  File "test/verify/check-machines", line *, in testBasic
+  File "test/verify/machineslib.py", line *, in testBasic
     wait(lambda: "Linux version" in self.machine.execute("cat {0}".format(args["logfile"])), delay=3)
 *
 Error: Condition did not become true.

Using libvirt D-Bus interface from https://github.com/libvirt/libvirt-dbus/ This commit is rewritting martinpitt@329e623 with the major difference that it's keeping both Libvirt (virsh) and LibvirtDBus providers. The new provider at this point has exactly the same behavior and no extra functionality compared to Libvirt provider. Note: libvirt-dbus should be used only with libvirt version > v4.3.0 or making sure that commit v4.2.0-104-g4d7384eb9d is included. All tests where moved to machineslib.py. Two new files with test classes were created for the two providers, both inheriting from the TestMachines class. Closes cockpit-project#9240

… in install

martinpitt · 2018-09-18T13:03:12Z

subVmTestCreate15 failure debugging notes:

This test case checks creating a VM with an unreasonable amount of memory, which should fail.
The screenshot shows one VM in state "in transition". After some time, this goes away, the list is empty in cockpit and in virsh list --all.
As per @KKoukiou , the "in transition" state should not happen at all, and it means that virt-install did not fail as it is supposed to.
While the b.wait_present("thead tr td") is still going on, logging into cockpit in a separate browser immediately shows an empty VM list. virsh list --all is also empty, and there are no qemu processes. So this looks just like the UI of the tested browser is getting confused. Could it be that the default action that sets "in transition" races with the failure of create_machine.sh, so that the latter is processed after the former?
I added set -x to create_machine.sh, and it indeed appears to not have failed: If handle_failure() were called, it should have an exit $?, which clearly does not appear. Instead, the last read FINISH_LINE appears and the rm -f at the bottom of the script.

+ VM_NAME=subVmTestCreate15
+ SOURCE=/var/lib/libvirt/novell.iso
+ OS=macosx10.5
+ MEMORY_SIZE=921600
+ STORAGE_SIZE=0.009765625
+ START_VM=false
++ awk 'BEGIN{ print "0.009765625"<=0 }'
+ COMPARISON=0
+ '[' 0 -eq 1 ']'
+ DISK_OPTIONS=size=0.009765625,format=qcow2
+ GRAPHICS_PARAM='--graphics vnc,listen=127.0.0.1'
+ spiceSupported
++ uname -m
+ ARCH=x86_64
+ case "$ARCH" in
+ QEMU=qemu-system-x86_64
+ type -P qemu-system-x86_64
+ grep -q '"enabled":'
+ qemu-system-x86_64 --qmp stdio --nographic -nodefaults
+ printf '{"execute":"qmp_capabilities"}\n{"execute":"query-spice"}\n{"execute":"quit"}'
+ GRAPHICS_PARAM='--graphics spice,listen=127.0.0.1 --graphics vnc,listen=127.0.0.1'
+ '[' macosx10.5 = other-os -o -z macosx10.5 ']'
++ cut -c 1
++ echo /var/lib/libvirt/novell.iso
+ FIRST_1_SOURCE=/
++ cut -c -3
++ echo /var/lib/libvirt/novell.iso
+ FIRST_3_SOURCE=/va
++ cut -c -4
++ echo /var/lib/libvirt/novell.iso
+ FIRST_4_SOURCE=/var
+ '[' false = true -a / = / -a '!' -f /var/lib/libvirt/novell.iso ']'
+ '[' false = true -a '(' / = / -o /var = http -o /va = ftp -o /va = nfs ')' ']'
+ LOCATION_PARAM=
++ mktemp
+ XMLS_FILE=/tmp/tmp.lXP5thC14u
+ '[' false = true ']'
+ STARTUP_PARAMS=--print-xml
+ HAS_INSTALL_PHASE=true
+ virt-install --name subVmTestCreate15 --os-variant macosx10.5 --memory 921600 --quiet --disk size=0.009765625,format=qcow2 --print-xml --graphics spice,listen=127.0.0.1 --graphics vnc,listen=127.0.0.1
+ '[' false = true ']'
++ grep -n '</domain>' /tmp/tmp.lXP5thC14u
+ DOMAIN_MATCHES='63:</domain>'
++ wc -l
++ echo '63:</domain>'
+ LAST_STEP=1
+ CURRENT_STEP=1
+ START_LINE=1
+ read -r FINISH_LINE
+ sed 's/[^0-9]//g'
+ echo '63:</domain>'
++ expr 63 + 1
+ QUIT_LINE=64
+ '[' 1 = 1 ']'
+ sed -n -i '1,63p;64q' /tmp/tmp.lXP5thC14u
++ sed 's/[^0-9]//g'
++ grep -n '</metadata>' /tmp/tmp.lXP5thC14u
+ METADATA_LINE=
+ METADATA='    <cockpit_machines:data xmlns:cockpit_machines="https://github.com/cockpit-project/cockpit/tree/master/pkg/machines"> \
      <cockpit_machines:has_install_phase>true</cockpit_machines:has_install_phase> \
      <cockpit_machines:install_source>/var/lib/libvirt/novell.iso</cockpit_machines:install_source> \
      <cockpit_machines:os_variant>macosx10.5</cockpit_machines:os_variant> \
    </cockpit_machines:data>'
+ '[' -z '' ']'
++ wc -l
++ cat /tmp/tmp.lXP5thC14u
+ METADATA_LINE=63
+ METADATA='\ \ <metadata> \
    <cockpit_machines:data xmlns:cockpit_machines="https://github.com/cockpit-project/cockpit/tree/master/pkg/machines"> \
      <cockpit_machines:has_install_phase>true</cockpit_machines:has_install_phase> \
      <cockpit_machines:install_source>/var/lib/libvirt/novell.iso</cockpit_machines:install_source> \
      <cockpit_machines:os_variant>macosx10.5</cockpit_machines:os_variant> \
    </cockpit_machines:data> \
  </metadata>'
+ virsh -q define /dev/stdin
+ sed '63i \ \ <metadata> \
    <cockpit_machines:data xmlns:cockpit_machines="https://github.com/cockpit-project/cockpit/tree/master/pkg/machines"> \
      <cockpit_machines:has_install_phase>true</cockpit_machines:has_install_phase> \
      <cockpit_machines:install_source>/var/lib/libvirt/novell.iso</cockpit_machines:install_source> \
      <cockpit_machines:os_variant>macosx10.5</cockpit_machines:os_variant> \
    </cockpit_machines:data> \
  </metadata>' /tmp/tmp.lXP5thC14u
+ read -r FINISH_LINE
+ rm -f /tmp/tmp.lXP5thC14u

warning: GetXMLDesc method failed for path /org/libvirt/QEMU/domain/_ae181dd8_d548_4f7f_88b9_dde6b5ac1b7f 
warning: ERROR    internal error: qemu unexpectedly closed the monitor: 2018-09-18T12:56:12.128864Z qemu-system-x86_64: cannot set up guest memory 'pc.ram': Cannot allocate memory
Domain installation does not appear to have been successful.
If it was, you can restart your domain by running:
  virsh --connect qemu:///session start subVmTestCreate15
otherwise, please restart your installation.

martinpitt · 2018-09-18T13:59:24Z

I now added set + to install_machine.sh, which is the one that is supposed to fail. And it does:

warning: GetXMLDesc method failed for path /org/libvirt/QEMU/domain/_2998e19d_220f_4d9c_8be8_b473b0dbdd58 
warning: + VM_NAME=subVmTestCreate15
+ SOURCE=/var/lib/libvirt/novell.iso
+ OS=macosx10.5
+ MEMORY_SIZE=921600
+ VCPUS=1
+ DISKS='path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2
'
+ DISPLAYS='type=spice,listen=127.0.0.1
type=vnc,listen=127.0.0.1,port=-1

'
+ '[' -z macosx10.5 ']'
+ '[' -z 'path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2
' ']'
+ createOptions 'path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2
' --disk
+ CREATE_OPTIONS_RESULT=
+ IFS=
+ read -r PARAM
+ '[' -n '    path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2' ']'
+ CREATE_OPTIONS_RESULT=' --disk     path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2'
+ IFS=
+ read -r PARAM
+ '[' -n '' ']'
+ IFS=
+ read -r PARAM
+ DISKS_PARAM=' --disk     path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2'
+ '[' -z 'type=spice,listen=127.0.0.1
type=vnc,listen=127.0.0.1,port=-1

' ']'
+ createOptions 'type=spice,listen=127.0.0.1
type=vnc,listen=127.0.0.1,port=-1

' --graphics
+ CREATE_OPTIONS_RESULT=
+ IFS=
+ read -r PARAM
+ '[' -n '    type=spice,listen=127.0.0.1' ']'
+ CREATE_OPTIONS_RESULT=' --graphics     type=spice,listen=127.0.0.1'
+ IFS=
+ read -r PARAM
+ '[' -n type=vnc,listen=127.0.0.1,port=-1 ']'
+ CREATE_OPTIONS_RESULT=' --graphics     type=spice,listen=127.0.0.1 --graphics type=vnc,listen=127.0.0.1,port=-1'
+ IFS=
+ read -r PARAM
+ '[' -n '' ']'
+ IFS=
+ read -r PARAM
+ '[' -n '' ']'
+ IFS=
+ read -r PARAM
+ GRAPHICS_PARAM=' --graphics     type=spice,listen=127.0.0.1 --graphics type=vnc,listen=127.0.0.1,port=-1'
++ cut -c 1
++ echo /var/lib/libvirt/novell.iso
+ FIRST_1_SOURCE=/
++ cut -c -3
++ echo /var/lib/libvirt/novell.iso
+ FIRST_3_SOURCE=/va
++ cut -c -4
++ echo /var/lib/libvirt/novell.iso
+ FIRST_4_SOURCE=/var
+ '[' / = / -a '!' -f /var/lib/libvirt/novell.iso ']'
+ '[' / = / -o /var = http -o /va = ftp -o /va = nfs ']'
+ LOCATION_PARAM='--cdrom /var/lib/libvirt/novell.iso'
++ mktemp
+ DOMAIN_FILE=/tmp/tmp.z6auzHq3sj
+ virsh -q destroy subVmTestCreate15
+ virsh -q dumpxml subVmTestCreate15
+ virsh -q undefine subVmTestCreate15 --managed-save
+ virt-install --name subVmTestCreate15 --os-variant macosx10.5 --memory 921600 --vcpus 1 --quiet --wait -1 --noautoconsole --noreboot --disk path=/home/admin/.local/share/libvirt/images/subVmTestCreate15.qcow2,device=disk,bus=ide,readonly=off,shareable=off,driver_name=qemu,driver_type=qcow2 --cdrom /var/lib/libvirt/novell.iso --graphics type=spice,listen=127.0.0.1 --graphics type=vnc,listen=127.0.0.1,port=-1
ERROR    internal error: qemu unexpectedly closed the monitor: 2018-09-18T13:26:46.201467Z qemu-system-x86_64: cannot set up guest memory 'pc.ram': Cannot allocate memory
Domain installation does not appear to have been successful.
If it was, you can restart your domain by running:
  virsh --connect qemu:///session start subVmTestCreate15
otherwise, please restart your installation.
+ EXIT_STATUS=1
+ '[' 1 -eq 0 ']'
+ vmExists subVmTestCreate15
+ grep -q --line-regexp --fixed-strings subVmTestCreate15
+ virsh list --all
+ awk '{print $2}'
+ virsh -q define /tmp/tmp.z6auzHq3sj
+ rm -f /tmp/tmp.z6auzHq3sj
+ exit 1

~~I noticed someting interesting in the journal:~~

Sep 18 09:28:07 localhost.localdomain systemd-coredump[3728]: Process 1640 (libvirtd) of user 1000 dumped core.

~~That might or might not be related. Are these tests covering system or user qemu?~~

Update: I got another failed test run without this crash, so it's unrelated.

…ally dummy

martinpitt · 2018-09-18T17:30:08Z

Not fully perfect yet, but this looks so much better, thanks!

martinpitt

... and it only took 165 conversation comments and 59 force-pushes!

🎉

Using libvirt D-Bus interface from https://github.com/libvirt/libvirt-dbus/ This commit is rewritting martinpitt@329e623 with the major difference that it's keeping both Libvirt (virsh) and LibvirtDBus providers. The new provider at this point has exactly the same behavior and no extra functionality compared to Libvirt provider. Note: libvirt-dbus should be used only with libvirt version > v4.3.0 or making sure that commit v4.2.0-104-g4d7384eb9d is included. All tests where moved to machineslib.py. Two new files with test classes were created for the two providers, both inheriting from the TestMachines class. Closes #9240

Using libvirt D-Bus interface from https://github.com/libvirt/libvirt-dbus/ This commit is rewritting martinpitt@329e623 with the major difference that it's keeping both Libvirt (virsh) and LibvirtDBus providers. The new provider at this point has exactly the same behavior and no extra functionality compared to Libvirt provider. Note: libvirt-dbus should be used only with libvirt version > v4.3.0 or making sure that commit v4.2.0-104-g4d7384eb9d is included. All tests where moved to machineslib.py. Two new files with test classes were created for the two providers, both inheriting from the TestMachines class. Closes cockpit-project#9240

phrdina reviewed May 25, 2018

View reviewed changes

pkg/machines/index.es6 Outdated Show resolved Hide resolved

mareklibra previously requested changes May 28, 2018

View reviewed changes

KKoukiou force-pushed the machines-dbus branch 3 times, most recently from bab41b4 to 8fa488e Compare May 30, 2018 11:04

KKoukiou force-pushed the machines-dbus branch 2 times, most recently from 94f2ef6 to 8611756 Compare June 14, 2018 12:29

KKoukiou force-pushed the machines-dbus branch from 8611756 to d4c76dc Compare June 14, 2018 12:35

KKoukiou force-pushed the machines-dbus branch from d4c76dc to 87876d3 Compare June 29, 2018 16:56

KKoukiou commented Jun 29, 2018

View reviewed changes

KKoukiou force-pushed the machines-dbus branch from 87876d3 to bfef692 Compare July 13, 2018 11:24

martinpitt mentioned this pull request Jul 17, 2018

machines: Create and Attach disk to a virtual machine #9166

Merged

5 tasks

KKoukiou force-pushed the machines-dbus branch 2 times, most recently from 6a2d70b to 5a151cb Compare July 19, 2018 08:16

KKoukiou mentioned this pull request Jul 19, 2018

machines: Move reusable code to libvirt-common.es6 #9692

Closed

martinpitt added the needs-rebase label Jul 20, 2018

KKoukiou force-pushed the machines-dbus branch from 5a151cb to 2a67b7b Compare July 20, 2018 10:22

martinpitt added the needswork label Jul 20, 2018

martinpitt changed the title ~~RFC: machines: Add LibvirtDBus provider~~ WIP: RFC: machines: Add LibvirtDBus provider Jul 20, 2018

KKoukiou force-pushed the machines-dbus branch from 2a67b7b to 009f9f3 Compare July 20, 2018 13:24

KKoukiou force-pushed the machines-dbus branch from 009f9f3 to 42c41eb Compare July 24, 2018 18:49

phrdina suggested changes Jul 27, 2018

View reviewed changes

KKoukiou force-pushed the machines-dbus branch 4 times, most recently from 674214d to db27c9a Compare August 8, 2018 14:13

KKoukiou added 4 commits September 17, 2018 21:00

machines: Move libvirt.es6 file to libvirt-virsh.es6

015dabd

A new provider is going to be introduced in followup patch and we need to clarify from the file name the provider used in each case.

machines: actions.es6: Add id parameter to specific actions

df639bb

The id parameter is going to be used by followup changes introducing new provider.

machines: libvirt-common: Remove jquery dependency

8b98b25

Replace parseXML from jquery with DOMParser.

KKoukiou force-pushed the machines-dbus branch from 7783c08 to eb76890 Compare September 17, 2018 19:01

martinpitt removed the needs-rebase label Sep 17, 2018

KKoukiou force-pushed the machines-dbus branch from eb76890 to 63a3c7c Compare September 17, 2018 22:08

KKoukiou added 3 commits September 18, 2018 13:58

machines: Fix dom element id when disk stats are not supported

46dbb82

machines: testCreate: Click explicitely on overview tab after failure…

935dedd

… in install

KKoukiou force-pushed the machines-dbus branch from 63a3c7c to 935dedd Compare September 18, 2018 12:02

machines: UPDATE_UI_VM reducer should make sure that updated VM is re…

378e000

…ally dummy

martinpitt approved these changes Sep 18, 2018

View reviewed changes

martinpitt merged commit 1034b8f into cockpit-project:master Sep 18, 2018

KKoukiou mentioned this pull request Aug 8, 2018

machines: Detach disk #9812

Merged

KKoukiou deleted the machines-dbus branch November 29, 2018 10:58

KKoukiou mentioned this pull request Apr 6, 2021

machines: libvirt D-Bus provider: Add EventListener for close signal on libvirt-dbus service cockpit-project/cockpit-machines#75

Closed

machines: Add LibvirtDBus provider #9240

machines: Add LibvirtDBus provider #9240

Conversation

KKoukiou commented May 25, 2018 • edited by martinpitt Loading

mareklibra left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

KKoukiou commented Jun 14, 2018

KKoukiou commented Jun 29, 2018

Choose a reason for hiding this comment

martinpitt commented Jul 20, 2018

KKoukiou commented Jul 20, 2018

phrdina left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

KKoukiou commented Sep 17, 2018

martinpitt commented Sep 17, 2018

KKoukiou commented Sep 17, 2018

martinpitt commented Sep 18, 2018

martinpitt commented Sep 18, 2018

KKoukiou commented Sep 18, 2018

martinpitt commented Sep 18, 2018

martinpitt commented Sep 18, 2018 • edited Loading

martinpitt commented Sep 18, 2018

martinpitt left a comment

Choose a reason for hiding this comment

KKoukiou commented May 25, 2018 •

edited by martinpitt

Loading

martinpitt commented Sep 18, 2018 •

edited

Loading