We forget to change home for posix user when we renamed
'sems' to 'ngcp-sems' somewhere in ~mr3.2
It is too late to change it now. Also there is no big deal to have
home pointed to non-existed folder as nobody login under user sems.
Change-Id: I59c722150f7156b07f8a625dae9a72fe8da27e42
(cherry picked from commit 424bd26c5b)
On Debian/stretch we get /nonexistent as the home
for user mysql, while on Debian/jessie it's /var/lib/mysql.
Backported to mr4.5 as systems installed on 2.8 and
upgraded to mr3.8 and then to mr4.5 has '/nonexistent'
due to Percona DB is being used on 2.8
Change-Id: I87d6db988c5d973d5ef814a89e658080e0bd7e50
(cherry picked from commit 9c2db0284a)
Freshly installed systems use IDs 10000/10001 nowadays,
while systems that are upgraded from old(er) releases use IDs 1000/1001.
It happens because in the past users sipwise/cdrexport were
created by installer directly, while currently ngcp-support-access
cares about sipwise/cdrexport user using new id based 10000+
Change-Id: I1a08aa722d1648ebef451c111dea4ff8750a54eb
(cherry picked from commit 284abaa49b)
ngcp-panel/rest-api is very slow to start on
servers builder[1,3,4].mgm
Change-Id: I639b79588a7aedd0879f839d5c5c79f40deccf41
(cherry picked from commit f446d5cb8a)
Binary built via:
cd $GOPATH/src/github.com/aelsabbahy/goss/
CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -ldflags "-X main.version=mika~git804995a8_with_tap_and_junit" -o release/goss-linux-amd64 github.com/aelsabbahy/goss/cmd/goss
using goss.git based on 804995a8 and including my PRs
https://github.com/aelsabbahy/goss/pull/48 (junit support) +
https://github.com/aelsabbahy/goss/pull/47 (TAP support)
Also use /usr/sbin/goss if it's present and only if it's not
yet there download the according version. This makes
development and testing of custom goss versions easier.
While at it fix string<->integer issues like:
| DEPRICATION WARNING: false: command.exit-status should be an integer not a string
Change-Id: I3dadc0b604db565b11adf1839b11e29437856fd1
The Jenkins test 'netinstall-test-ngcp-installer' restores plain Debian VM
and install NGCP using ngcp-installer only.
In this case the test should skip some strict host/names checks.
Change-Id: I9b84784ebed0a54ff51aa2baef274f5466879352
Checking only for *.tt2 files isn't enough, we also have
*.services + *.postbuild files and want to also catch
any files that don't match our file suffix.
Change-Id: Ia7477ac45b4fc9e1ab3cc0188c35424bffeceedb
Check for getty process without path information and arguments,
because on systemd powered systems we have:
| /sbin/agetty tty1 38400
while sysvinit powered systemd run:
| /sbin/getty 38400 tty1
| /sbin/getty 38400 tty2
| /sbin/getty 38400 tty3
| /sbin/getty 38400 tty4
| /sbin/getty 38400 tty5
| /sbin/getty 38400 tty6
As soon as one "getty" process is available that should be fine enough
for us, no matter whether it's /sbin/agetty or /sbin/getty.
(Of course it would be even better if we'd have regex support
in the tests, but that's not available in upstream code...)
We are not going to check exect tty numer avialability,
otherwise test will failed in case is someone loged in to
this console. Lets check at least some terminal avialability.
In failing tests it's hard to parse very long cmdlines, so reduce
it as much as possible and still sensible. Also fix quoting of
the heartbeat command line for YAML format.
Alexander Lutay
Michael Prokop
Kirill Solomko
Victor Seva