MongoDB RPM packages (provided by official repo.mongodb.org repository, as of version 3.4.1) automatically and unconditionally restart server on package upgrade. This behaviour is hardcoded into postun handler:
if test $1 -ge 1
then
/usr/bin/systemctl restart mongod >/dev/null 2>&1 || :
fi
This is an inconvenient and dangerous behaviour, especially when you use configuration management tools to set up your servers. For example, I'd like to run a full Ansible playbook to set up my servers first, and then restart MongoDB manually one by one to have full control of the situation.
Is there any way to alter or disable this? Alternative MongoDB packages, maybe? Or some obscure yum/rpm command option to disable scriptlets?
I'm aware that I can switch to simple .tar.gz installation, but this is the last resort.
If you first download the rpm and install it manually using rpm; you can use the --nopostun option:
rpm -Uvh mongodb***rpm --nopostun
from the rpm man page:
--noscripts
--nopre
--nopost
--nopreun
--nopostun
--nopretrans
--noposttrans
Don't execute the scriptlet of the same name. The --noscripts option is equivalent to
--nopre --nopost --nopreun --nopostun --nopretrans --oposttrans
and turns off the execution of the corresponding %pre, %post, %preun, %postun %pretrans, and %posttrans scriptlet(s).
afaik yumcannot handle the --nopostun and other flags.
Related
I would like to use tox to run py.test on a project which needs additional setup in addition to installing packages into the virtualenv. After creating the virtualenv and installing dependencies, some commands need to be run.
Specifically I'm talking about setting up a node and npm environment using nodeenv:
nodeenv --prebuilt -p
I see that tox allows me to provide a custom command used for installing dependencies by setting install_command in tox.ini. But I don't think this is what I want because that replaces the command (I assume pip) used to install dependencies.
I thought about using a py.test fixture with session scope to handle setting up nodeenv but that seems hacky to me as I don't want this to happen when py.test is run directly, not via tox.
What is the least insane way of achieving this?
You can do all necessary setup after the creation of the virtualenv and the dependency installation in commands. Yes, it says "the commands to be called for testing." but if you need to do extra work to prepare for testing you can just do it right there.
It works through whatever you throw at it in the order it is given - e.g.:
[testenv:someenv]
deps =
nodeenv
pytest
flexmock
commands =
nodeenv --prebuilt -p
; ... and whatever else you might need to do
py.test path/to/my/tests
If you have commands/scripts or whatever else that produces the right result but it returns a non zero exit status you can ignore that by prepending - (as in - naughty-command).
If you need more steps to happen you can wrap them in a little (Python) script and call that script instead as outlined in https://stackoverflow.com/a/47834447/2626627.
There is also an issue to add the ability to use more than one install command: https://github.com/tox-dev/tox/issues/715 is implemented.
I had the same issue, and as it was important for me to be able to create the environment without invoking the tests (via --notest), I wanted the install to happen in the install phase and not the run phase, so I did something slightly differently. First, I created a create-env script:
#!/usr/bin/env sh
set -e
pip install $#
nodeenv --prebuilt --python-virtualenv --node=8.2.1
Made it executable, Then in tox.ini:
[tox]
skipsdist = True
[testenv]
install_command = ./create-env {opts} {packages}
deps = nodeenv
commands = node --version
This complete example runs and outputs the following:
$ tox
python create: .../.tox/python
python installdeps: nodeenv
python installed: nodeenv==1.3.0
python runtests: PYTHONHASHSEED='1150209523'
python runtests: commands[0] | node --version
v8.2.1
_____________________________________________________________________ summary ______________________________________________________________________
python: commands succeeded
congratulations :)
This approach has the downside that it would only work on Unix.
In tox 715, I propose the possibility of native support for multiple install commands.
I am using OpenERP 7.0-20130811-231021 in Ubuntu. I developed a new module but when I try to install it, that module is not in the OpenERP module list. I tried the following steps:
sudo /etc/init.d/openerp start
update modules list
go to installed module and viewed the module.
It is not there when I tried this command:
sudo ./openerp-server -u modulename
It shows the following:
error: [Errno 98] Address already in use
What is causing this error and how can I fix it?
Find openerp-server process id:-
ps -ax | grep openerp-server this will give the process id and need to kill it for example 1234
Kill Process ID:-
sudo kill -9 1234
Start Sever:-
sudo /etc/init.d/openerp-server start
And update module from the GUI, Hope this will help you.
In order to see a custom module in OpenERP 7, it must first be in the addons directory.
Go to Settings > Modules > Update Modules List Click Update You must have Technical Features enabled for the user you are logged in as.
Then go to Settings > Modules > Installed Modules Remove the [Installed] filter and search for your custom module
Custom modules will not appear in Settings > Modules > Apps because that view will only display Modules/Apps that are found online.
You need to check:
1- Your openerp-server.conf in install file, in there you should check that your addons_path reference your module location
2-Check your openerp.py and make shure everything is rigth
3- You must check also your Run Configurations Parameters and specified -u and module name, this will update your module in every restart
I'm trying to set up a CoffeeScript build system in Sublime Text 3, but I keep getting the following error:
env: node: No such file or directory
[Finished in 0.0s with exit code 127]
[cmd: ['coffee', '-o','/Users/jcourtdemone/Sites/autotempest.com/new_design_sandbox/static/script', '-cw', '/Users/jcourtdemone/Sites/autotempest.com/new_design_sandbox/static/coffee']]
[dir: /Users/jcourtdemone/Sites/autotempest.com/new_design_sandbox/static/coffee]
[path: /usr/bin:/bin:/usr/sbin:/sbin]
My build system looks like this:
{
"name": "Coffee - AT",
"cmd": ["coffee","-o","${project_path:${folder}}/static/script","-cw","${project_path:${folder}}/static/coffee"],
"selector": "source.coffee",
"path":"/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/lib/node_modules/coffee-script/bin"
}
Two things strange about this.
1) It says it's looking in /usr/bin where a symlink to coffee exists.
2) Because of (1), I overrode $PATH to include the actual location of coffee which is /usr/local/lib/node_modules/coffee-script/bin, but for some reason, $PATH isn't being overridden properly, it's sticking with the default $PATH.
Things to note:
i) I've verified that all paths are correct and pass normally through a regular terminal command.
ii) Tried with a "shell": true variable in the build system.
iii) I have another build system for Compass like this that works fine.
Anyone run into similar problems or issues? Any ideas?
In Terminal, type which node, then create a symlink to that location in /usr/bin. For example, if node lives in /usr/local/bin, create the symlink like so:
sudo ln -s /usr/local/bin/node /usr/bin/node
If you look at the source of your coffee script, you'll probably find that the first line is something along the lines of:
#!/usr/bin/env node
Exit code 127 in Sublime means that an env command has failed - so in your case, the build system is finding coffee, but it can't execute it because the node binary isn't in Sublime's default search path.
There are two ways to redefine the default search path for Sublime. The first (and easiest) is to always open it from the command line using the built-in subl command. If you're an OS X power user and don't mind messing with important system settings, check out my post on unix.SE on how to alter the default /usr/bin:/bin:/usr/sbin:/sbin path that you're seeing. Be forewarned that if you don't do things correctly, you may break your system. However, if you're running Mountain Lion (10.8.X) and you follow the instructions exactly, everything should be fine. (I haven't upgraded to Mavericks, so no guarantees on whether it'll work with that version.)
How to solve the problem under an Ubuntu System
The fact is "coffee" command will call /usr/bin/node to continue its work, however, the original "node" command for the node application on an Ubuntu system is changed from "node" to "nodejs" to avoid name conflicting. That is the reason, the shell will compliant you "/usr/bin/env: node: No such file or directory". whenever you type
$ coffee
To solve the bug, just let the shell find something named "node" in its default searching path, and this so-called "node" will promote nodejs. The command "nodejs" lies under path of /usr/bin/nodejs.
We will use symbol link to link "node" with nodejs, and place the link "node" within the default searching path, so that the shell will find it.
sudo ln -s /usr/bin/nodejs /usr/bin/node
But beware, make sure that you do NOT have another "node" command under /usr/bin/, you can check it by try to run
$ which node
I do NOT know what to do if you have installed another "node" application.
In Ubuntu you can install the package nodejs-legacy
sudo apt-get install nodejs-legacy
this package just create a symbolic link to binary nodejs
You should be able to fix this all in your build system without needing to add a symlink on your machine.
For example if node lives in /usr/local/bin/node all you have to do is change the path in your build_system to be:
"path": "/usr/local/bin:$PATH"
I had the same problem with Sublime Text 2.
Creating this sublime build worked for me:
{
"cmd": ["coffee", "-c", "$file"],
"selector" : "source.coffee",
"path" : "/usr/local/lib/node_modules/coffee-script/bin/:/usr/local/bin:$PATH"
}
The following code worked for me in Ubuntu 14.04:
**$ sudo apt-get install NodeJS-legacy**
The other problem was the version checking frameworks such as for e.g: gulp -v the same code also solved this problem.
Type the next in the console:
sudo ln -s /usr/bin/nodejs /usr/bin/node
Normally in linux Debian we do sth like this to install a package non-interactively e.g
sudo apt-get install -y Package_x_z
#[-y --assume-yes]
How we can do the same while installing a perl module e.g
sudo perl -MCPAN -e 'install DBI'
That prompt is (typically) coming from ExtUtils::MakeMaker's prompt() function. Stick export PERL_MM_USE_DEFAULT=1 in your .bashrc (or equivalent for your preferred shell) to stop the prompts. The ExUtils::MakeMaker man page documents it thus:
PERL_MM_USE_DEFAULT
If set to a true value then MakeMaker's prompt function will always return the default
without waiting for user input.
Note that this can come to bite you if you run cpan(1) on a box that's not yet had CPAN repositories configured. It will rattle on and get stuck in a prompt loop at a point where there is no default and you need to make a choice, but have no ability to do so. export PERL_MM_USE_DEFAULT=0 in the shell before running cpan(1) will of course temporarily re-enable input.
To prevent the CPAN client from asking whether to install prerequisites, start it in interactive mode
perl -MCPAN -e shell
and enter the commands:
o conf build_requires_install_policy yes
o conf prerequisites_policy follow
o conf commit
The commit command is optional, but it will update the default configuration, which I suspect is what you want. Without it, you may or may not (depending on whether autocommit is enabled in your CPAN config) need to make this change every time you want to do a prompt-less installation.
These changes will deal with all of the CPAN client's routine questions about whether to install dependencies. For distributions which have questions embedded in their install scripts, you may also want to add
o conf inactivity_timeout 60
to set how long it will wait for a response before automatically going with the default answer to the question. (Set it to 0 to change it back to "wait forever".)
What about just :
$ yes | sudo perl -MCPAN -e 'install DBI'
Ban ! your problem is solved :-)
Appending to an answer here, you can also make these changes in config file located at /usr/share/perl5/CPAN/Config.pm.
'build_requires_install_policy' => q[yes],
'prerequisites_policy' => q[follow],
This helped me to automate installation, since CPAN doesn't have these configuration by default.
I have installed perlbrew which seems like a good solution, but I get some meaningless error when actually trying to install some Perl version:
$ perlbrew install perl-5.12.1
Attempting to load conf from /home/dave/perl5/perlbrew/Conf.pm
Fail to get http://search.cpan.org/dist/perl-5.12.1 (error: ) at /home/dave/perl5/perlbrew/bin/perlbrew line 1277.
Based on your comments do you have http_proxy ENV variable set in your shell?
$ env | grep http_proxy
If not then set it with your proxy settings and re-try perlbrew install:
$ export http_proxy = "http://yourProxyURLorIP:8080"
$ perlbrew install perl-5.12.1
perlbrew uses this ENV variable to pick up the proxy server. If this ENV variable isn't set then it tries the normal direct HTTP connection (see line 1274 in perlbrew current master on Github)
$ua->proxy($ENV{http_proxy}) if $ENV{http_proxy};
If that doesn't work then have a look at HTTP::Lite. This is what perlbrew uses under the hood to fetch source code. NB. perlbrew uses its own copy of HTTP::Lite
Finally if still no luck you mentioned that you "first installed it" via CPAN. The docs does mention issues when upgrading from a previous CPAN version. This maybe something further you need to look into?
Update Test this HTTP::Lite script and let me know what you see (NB. You may need to install HTTP::Lite):
use strict;
use warnings;
use HTTP::Lite;
my $ua = HTTP::Lite->new;
$ua->proxy("yourProxyURLorIP:8080"); # <= http_proxy env minus "http://"
my $req = $ua->request( 'http://search.cpan.org/dist/perl-5.12.1/' )
or die "Unable to get document: $!";
print $ua->body(); # <= if you get this then all is good!
I think you've probably been hit by a known bug with HTTP::Lite, see RT issue uri style proxy env vars fail to set the proxy and port correctly.
The above code is the workaround to this bug. I assume the same bug is in perlbrew copy of HTTP::Lite. If it is then removing the http:// from your http_proxy ENV would resolve the problem (famous last words!)
Update 2
Just to make my last comment clear when you run perlbrew you can do this (from shell like bash):
http_proxy=IPaddr:Port perlbrew install perl-5.12.1
You would need to always prefix every perlbrew command like this, at least until HTTP::Lite or perlbrew proxy bug is fixed.
Alternative to above is you can just patch your local version just be adding the following before line 1277:
$ENV{http_proxy} = "IPaddr:Port"; # <= your proxy IP & port
Hopefully we've finally cracked it! Let me know if all successful because if so then I'll post a fix to Gugod (author of perlbrew) with necessary local changes to HTTP::Lite.