Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
Closes: ARMBAND-53
Change-Id: I92c1541a22a837c281d9b86f6ee8da327305884d
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
In order to clean up old partitioning and boot entries on target
nodes, before removing the Fuel Master, try ssh-ing into it and
removing all environments/nodes.
This is especially important for EFI systems, where old boot
entries are left behind without a proper env delete, leading to
target nodes failing to PXE boot on Fuel Master re-install.
This change assumes that:
- all Fuel Master information is unchanged between old and new
deploy (fuel IP, password etc.);
- Fuel Master is up and running, in a sane state (target nodes
are also online), when deploy starts;
Closes: ARMBAND-51
Change-Id: I52b853b37a0d226a9cb33179f7aeb65eff74da4e
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
On EFI-enabled systemd, grub-install from grub-efi-* package
installs a boot entry named "ubuntu".
Since this boot entry is saved in board flash memory, erasing
the MBR bootloader code and/or partition signature will not
clear the ubuntu boot entry, leaving it pointing to a loader
on the ESP (EFI System Partition) that will try to load the
kernel/initrd from an erased partition.
In Fuel 8.0, the whole disk was erased, so the ubuntu EFI boot
entry was skipped due to missing ESP, while for Fuel 9.0 and
above we have to explicitly remove it to keep the system in a
sane state.
Note: efibootmgr is installed automatically on EFI systems
as a dependency of grub-efi-*.
Closes: ARMBAND-47
Change-Id: If715e960dfd96df807abb0e26f4b7604624ac408
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Merged upstream.
This reverts commit 31087865b416f0eab066377c76a0d580d3fd8bd1.
Change-Id: I60bd902207bb35e244e0e10646226d9f611bd68a
|
|
libvirt >= 1.3.0 installs additional services, including virtlogd,
which need to be started explicitly after install, as the Ubuntu UCA
package is currently broken.
Fuel code already carries a fix for this, but only for UCA libvirt.
Refactor the condition for the existing fix to check package version
instead of origin.
Closes-bug: https://jira.opnfv.org/browse/ARMBAND-37
Change-Id: I651ab48072ddcbfb88ce78a66c25cb524e442981
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Upstream bug report [1], proposed change backported from [2].
Temporary backport proposed fix until fuel-library merges it.
Currently net.ipv4.ip_nonlocal_bind is not set inside vrouter
namespace.
This can lead to deployment failure with p_dns error:
ocf-ns_dns: INFO: dnsmasq daemon is not running
ocf-ns_dns: ERROR: dnsmasq: failed to create listening socket for
10.109.5.1: Cannot assign requested address
ocf-ns_dns: ERROR: Error. dnsmasq daemon returned error 0.
We had very similar issue in the past:
https://bugs.launchpad.net/fuel/+bug/1500871
We need to add
ocf_run $RUN_IN_NS /sbin/sysctl -w net.ipv4.ip_nonlocal_bind=1
To get_ns() in ns_dns and ns_vrouter
[1] https://bugs.launchpad.net/fuel/+bug/1595957
[2] https://review.openstack.org/#/c/334415/
Change-Id: Id8d3a2e58de4dccc9a79536bd443294e085f70ff
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I493279d81b32a58b918f2b6caa0cf869a0caa674
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Change-Id: I220b364d2aecc5e08b3c90c451acf26e5d8ea0d7
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
Change-Id: Icbbba0bee68a8fe9de76dbf3c339e01da43fb3a2
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Previous change [1] introduced the addition of an iptables rule
that was supposed to allow SSH access on all ifaces (not only
admin iface) when additional ifaces are configured.
However, Fuel installer is flushing the rules after transplant
adds our SSH config, overwriting it.
Move iptables SSH config to post-install section, as standalone
script. In order to keep the same behavior, test ifcfg-eth0
(admin interface is expected to be called eth0 by convention)
for "DEFROUTE=no" and only whitelist SSH on all ifaces if it matches.
[1] https://gerrit.opnfv.org/gerrit/#/c/16571/
Change-Id: I086b75461daa62671cad10494fe34acfd77757ae
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Idbc45b4d1aeddf5450b756d23faf56c95fccc975
|
|
|
|
Change-Id: Iae462ebe89ff3bc351addd12a8f5cc5ecdef2b11
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Change-Id: I2cd1899d82878dfece22ed546e1ff24d5387e8c1
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Previously (in Fuel 8.0), SSH used to listen only on 10.20.0.2
(admin interface), which required editing sshd_config and restarting
SSH server for allowing SSH connections over the public IP on eth1
(just an example, which corresponds to Armband PODs use cases).
In Fuel 9.0, SSH server on Fuel Master listens on all ifaces,
but connections are filtered by iptables.
This change piggy-backs on a previous Armband addition that allows
transplant.py script to configure additional interfaces (e.g. public).
In case additional interfaces have been configured by transplant,
this change will instruct iptables to accept SSH connections on ANY
interface, not only the admin one.
Possible improvements:
- limit SSH access to admin + interfaces added by transplant instead
of ANY;
Change-Id: I0923496e1b23f6dc826c8afbbe9805956c2d4b34
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
FIXME: Revert this when upstream Fuel 9.1 closes bug [1].
[1] https://bugs.launchpad.net/fuel/+bug/1597775
Change-Id: I692b0ef63ee3e035deadef9d4e47b43f4c53e4aa
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
For some reason, our docker sandbox builder already has dh-systemd
when checking for "fuel-agent" build-deps, and therefore skips
adding it as a build dep inside the chroot, leading to debuild fail.
Backporte upstream fix for this from [1].
[1] https://review.openstack.org/#/c/325210/
Change-Id: Ieece9b2250bd91e688bd0f203240da5234930558
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
My initial goal was to split the ODL patch in two patches:
- leveldb + leveldbjni arm64 support;
- opnfv-quagga building/packing for multiple archs (UBUNTU_ARCH);
However, due to the design of opnfv-packaging repo, it is easier to:
- patch upstream sources to support native arm64 builds (see [1]);
- add prebuilt arm64 binaries to output dir in git repo (like amd64),
submitted as pull request upstream [2];
Until above OPNFV-QUAGGA changes land upstream or are refactored,
rely on forked repository that contains above patches [3] on
branch "stable/colorado".
opnfv-quagga patch adding UBUNTU_ARCH parsing is more or less there
to detect missing DEBs, rather than filter prebuilt DEBs, which for
now only cover "amd64 arm64".
[1] https://github.com/nikolas-hermanns/opnfv-quagga-packaging/pull/1
[2] https://github.com/nikolas-hermanns/opnfv-quagga-packaging/pull/2
[3] https://github.com/alexandruavadanii/opnfv-quagga-packaging
Change-Id: I022c1e82531d82422486f2041c70a3f50b6ca386
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Iea330d6d778ff5034e1e801d20fa74485e302a95
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
Fuel@OPNFV just added the Yardstick plugin.
This plugin is a little different, instead of breaking fuel plugin
code apart from the contents (i.e. fuel-plugin-yardstick separate
from yardstick itself), it holds both fuel plugin metadata and
actual yardstick code in the same repo.
So far, yardstick plugin code itself is arch-independant and should
not be touched.
TODO:
- add yardstick patches for arm64 until we can validate Yardstick;
- submit patches added locally in Armband to upstream
(Yardstick@OPNFV);
Change-Id: I24bc8858b92c9bdd529a64de833d7b98c06a6ade
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Upstream change [1] required a rebase of our patch and
also transitioning leveldb build to using JAVA8.
FIXME: Transform JAVA_HOME hardcodes into something like (see [1]):
JAVA_HOME=$(readlink -f /usr/bin/java | sed "s:/jre/bin/java::")
https://gerrit.opnfv.org/gerrit/#/c/16091/
Change-Id: I7ea6e3989e575754a389115d44cc44e4c416b4d9
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
FIXME: dpdk, ovs, qemu etc. building only supports x86 for now.
TODO: conf dir should be patched too.
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I986bfb5ae549321a0696f945d24ccb1b6b198f96
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
Move fuel-main patching to opnfv-fuel, preparing for upstreaming
this to Fuel@OPNFV.
Drop fuel-main submodule!
Change-Id: I12b995f37701658dca1d1a212ede029d0e03231c
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
This was an interesting bug. The code has been broken for a while,
but we did not notice because nova was installed before these
puppets were ran.
With the Mitaka reworks and changes in the puppet manifests, this
bug surfaced, but manifested as `virtlogd` service not being
started, looking like a libvirt 1.3+ problem.
So, just add the missing "->" in the troubling puppet file.
Change-Id: Ic702e17e8f732bf7a277714cc4416e0f2ee68cfe
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Rebase conflicting patches due to format change.
Change-Id: Ia9d1f9c0cad77da7b95e8ec4b62324953ef3ce9c
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
Change-Id: Ibd41229ce744256fc289756708d64c3fedede9f4
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Previously, we added and used qemu-user-static and its wrapper
over debootstrap (qemu-debootstrap) in:
- requirements-rpm.txt
No longer needed, qemu-user-static should be added automatically
as a cobbler rpm dependency (similar to cobbler-grub-aarch64);
- prepare-build-env.sh
No longer needed, we only support building packages natively.
We never built an arm64 chroot at ISO build time.
- sandbox.mk
Same motivation as above, no longer needed.
I think this would have needed qemu-user-static on build machine,
and binfmt properly set up on that machine as well, adding
complexity to the build process, although we do not use it.
So, simplify things a little and prepare this for upstream.
Change-Id: Idfce5aa594f1fcd92b1f238c7d07cb88f26064b8
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Skip parsing comment lines in ifcfg-ethX files at reap, e.g.:
Change-Id: Ide6b2111619f7558f1434fda5cd254b842cef0ae
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
While at it split another patch in f-l in two.
Change-Id: Ife31a94447176798c3cf7052408ffeddb1e753ff
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I9442f217d2f840382b40f6eae77ddb9ae2ddbadc
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
|
|
Change-Id: I4f029b00b03c665bef2bbf367953e9d95ed215f1
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I3702648583983ac7afcfef7a9556fcd33e984457
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
I messed up the identation and missed a semicolon during rebase.
Change-Id: I7142721cc57115a72c93bcfd270c03d6cb04c399
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I3c7bd7f9be85078783a2e95dfe480cffc08534d1
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Upstream change "Improve reliability of upload_cirros.rb" [1]
allows us to drop the "Increase upload_cirros timeout" patch,
and also requires rebasing our direct kernel boot patch.
Also squashed in a small context edit for an unrelated patch.
[1] https://review.openstack.org/#/c/327792/
Change-Id: I8adc30a74714414fae55bc7b81bc0f69f41735ce
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
|