-
Notifications
You must be signed in to change notification settings - Fork 577
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Initialize Configuration#Concurrency from /sys/fs/cgroup/... #7845
base: master
Are you sure you want to change the base?
Conversation
|
TODO
|
... e.g. |
5793b4e
to
66b84be
Compare
MacOS
CentOS
Docker
|
@cla-bot check |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Doesn't work with cgroup2, there the information is in /sys/fs/cgroup/cpu*
files (i.e. not in extra subdirectories).
In general, an overview of how the Docker options map to cgroup attributes (and therefore which of these options this PR takes into account) would be nice. Looking at the corresponding systemd options is probably also interesting and they do properly document the cgroup options affected.
int Configuration::Concurrency{static_cast<int>(std::thread::hardware_concurrency())}; | ||
|
||
#ifdef __linux__ | ||
static std::string ReadSysLine(const char *file) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This function looks overly complex for what it does. Shouldn't this basically be the same as std::ifstream
+ std::getline
? The only thing I see that wouldn't be possible with that is the check for ENOENT
. If this is just for kernel version compatibility, using an extra check would be fine. If this file can (dis)appear at runtime (changing the cgroup options, moving the process between cgroups), have a look at boost::iostreams
like it's already used in AtomicFile
.
lib/base/configuration.cpp
Outdated
return ""; | ||
} | ||
|
||
throw system_error(error_code(errno, system_category())); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we really want to fail starting in this case? I think logging a warning and continuing with the value from std::thread::hardware_concurrency()
would be a better option.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- In "this case" a kernel subsystem(!) is out of service. The best reason to panic.
- Logging? We're in a static constructor.
|
|
TODO @Al2Klimov
|
66b84be
to
1d6ab00
Compare
Doesn’t whoever controls the container spec also control the command it runs? I.e. can't they just set -DConfiguration.Concurrency= in addition to --cpu-whatever? I.e. do we need this PR? |
Most likely yes.
Not strictly needed as it can be worked around. However, the user has to be aware of this and it would be nice if Icinga would just do the correct thing by itself. |
1d6ab00
to
cda810d
Compare
... not to try to consume more hardware resources than the admin assigned.
fixes #7842