During a DOR (Dead Office Recovery) procedure, the worker nodes are
taking longer than expected to become online, if the cluster network
is IPv6. This is related to an extra reboot occurring during the
bring-up process. Today only one DHCPv6 attempt is done, and, as the
controllers are still booting up, no response is available.
Eventually, this lack of a management IPv6 address will lead to an
extra reboot.
The correction is to make the DHCPv6 client continuously attempt to
get a management IP address from the server. If the management
network is IPv4, the DHCPv6 client is not called and will not
trigger continuous DHCPv6 requests.
Closes-bug: 1934676
Signed-off-by: Andre Fernando Zanella Kantek <AndreFernandoZanella.Kantek@windriver.com>
Change-Id: Ife5b06c5294e0014918073e9f6ee8bd372b85a78
This commit rebases initscripts patch set, dropping
run-dhclient-as-daemon-for-ipv6.patch
Currently, ifup-eth tries running ipv6 dhclient with the one-shot
option, and if that fails, then retries indefinitely in the background.
That has the side-effect of causing the ifup-post script to not be run
if the first dhclient attempt fails, which will prevent routes on that
interface from being created. This is especially problematic in the
case of a DOR, where the compute nodes may come up before dnsmasq is up
on the controller.
This is different from upstream centos, which will only try running
dhclient with the one-shot option for ipv6.
By reverting the initscripts patch to run as a daemon, ipv6 dhclient now
runs as one-shot only, and if it fails, ifup-eth script exits without
getting an address, and then the node fails to come up and reboot.
While this may result in the compute node having an extra reboot in a DOR,
that is preferable to the compute coming up incorrectly and requiring a
lock/unlock to recover.
Closes-bug: 1844579
Change-Id: I5b7f6b7c878dc4e4737d986f11fae3301585fb1c
Signed-off-by: Joseph Richard <joseph.richard@windriver.com>
The ifup-aliases script assumes that the IPv4 address is always
defined. If the configuration is only for IPv6, the script would
generate an error and not process the IPv6 address of the interface.
This commit is to bring up the IPv6 interface even if the IPv4 address
is not defined.
Partial-Bug: 1834234
Change-Id: Ib0c4cbc7ec19cc0c0c485e4ad63c380aa8a49a4c
Signed-off-by: Teresa Ho <teresa.ho@windriver.com>
The management interface uses the loopback interface for AIO-SX.
For the containerized deployment, the cluster-host interface shares
the same interface with management by default. It would need to be
scoped to host instead of global as it is an alias to the loopback
interface.
The scope parameter is added for the ifup-aliases to set scope
to host if the parent device is the loopback interface.
Story: 2004273
Task: 28590
Change-Id: Iee62c825ed1650b6cfdb0d4e5eb43b4a635e819a
Signed-off-by: Teresa Ho <teresa.ho@windriver.com>
Use initscripts-config package to package script and service file for
initscripts package.
Refactor 0001-Disable-zeroconf-route.patch,
spec-add-mountnfs-init-script.patch and
spec-include-TiS-changes.patch, let initscripts-config to be
responsible for the installation of config/script/service files.
Merged some meta patches that just includes adding source code patch to
spec-include-Tis-changes.patch.
Removed base/procps since it just includes one file, sysctl.conf. And
move this file to initscripts-config folder.The monitor-tools package
in stx-integ has a %post script that is adding an entry to sysctl.conf,
so add "Requires: initscripts-config" in monitor-tools.spec, to ensure
it is installed after this package replaces the file.
Deployment test and ping test between VMs pass
Service, config and script files check pass.
Story: 2003768
Task: 27585
Change-Id: I2ea3bd05bdc5bca5658d157e6f40f7380e922500
Signed-off-by: zhipengl <zhipengs.liu@intel.com>
Problem:
- Centos 7.5 upgraded initscripts.
- Porting of initscripts patches did not resolve and 'fuzz' in the line
numbers of the patches.
- If initscript is built by rpm 4.11, or default version of rpm
until 4.14 is compiled, a fuzzy patch results in the creating
of an .orig file.
- Packaging of initscripts failes due to the unexpected, and
unpackaged .orig file
Solution:
Safest solution is to de-fuzz our initscripts patches.
A seperate investigation will bug report will address the
broader issues.
Closes-Bug: 1794611
Change-Id: I1e91a230194c0c21798b7e91b444f929b54106d6
Signed-off-by: Scott Little <scott.little@windriver.com>