##// END OF EJS Templates
merge: mark file gets as not thread safe (issue5933)...
merge: mark file gets as not thread safe (issue5933) In default installs, this has the effect of disabling the thread-based worker on Windows when manifesting files in the working directory. My measurements have shown that with revlog-based repositories, Mercurial spends a lot of CPU time in revlog code resolving file data. This ends up incurring a lot of context switching across threads and slows down `hg update` operations when going from an empty working directory to the tip of the repo. On mozilla-unified (246,351 files) on an i7-6700K (4+4 CPUs): before: 487s wall after: 360s wall (equivalent to worker.enabled=false) cpus=2: 379s wall Even with only 2 threads, the thread pool is still slower. The introduction of the thread-based worker (02b36e860e0b) states that it resulted in a "~50%" speedup for `hg sparse --enable-profile` and `hg sparse --disable-profile`. This disagrees with my measurement above. I theorize a few reasons for this: 1) Removal of files from the working directory is I/O - not CPU - bound and should benefit from a thread pool (unless I/O is insanely fast and the GIL release is near instantaneous). So tests like `hg sparse --enable-profile` may exercise deletion throughput and aren't good benchmarks for worker tasks that are CPU heavy. 2) The patch was authored by someone at Facebook. The results were likely measured against a repository using remotefilelog. And I believe that revision retrieval during working directory updates with remotefilelog will often use a remote store, thus being I/O and not CPU bound. This probably resulted in an overstated performance gain. Since there appears to be a need to enable the thread-based worker with some stores, I've made the flagging of file gets as thread safe configurable. I've made it experimental because I don't want to formalize a boolean flag for this option and because this attribute is best captured against the store implementation. But we don't have a proper store API for this yet. I'd rather cross this bridge later. It is possible there are revlog-based repositories that do benefit from a thread-based worker. I didn't do very comprehensive testing. If there are, we may want to devise a more proper algorithm for whether to use the thread-based worker, including possibly config options to limit the number of threads to use. But until I see evidence that justifies complexity, simplicity wins. Differential Revision: https://phab.mercurial-scm.org/D3963
Gregory Szorc -
r38755:be498426 default
Show More
Name Size Modified Last Commit Author
/ contrib / docker / apache-server
Dockerfile Loading ...
README.rst Loading ...
entrypoint.sh Loading ...
hgwebconfig Loading ...
vhost.conf Loading ...

Apache Docker Server

This directory contains code for running a Mercurial hgweb server via mod_wsgi with the Apache HTTP Server inside a Docker container.

Important

This container is intended for testing purposes only: it is not meant to be suitable for production use.

Building Image

The first step is to build a Docker image containing Apache and mod_wsgi:

$ docker build -t hg-apache .

Important

You should rebuild the image whenever the content of this directory changes. Rebuilding after pulling or when you haven't run the container in a while is typically a good idea.

Running the Server

To run the container, you'll execute something like:

$ docker run --rm -it -v `pwd`/../../..:/var/hg/source -p 8000:80 hg-apache

If you aren't a Docker expert:

  • --rm will remove the container when it stops (so it doesn't clutter your system)
  • -i will launch the container in interactive mode so stdin is attached
  • -t will allocate a pseudo TTY
  • -v src:dst will mount the host filesystem at src into dst in the container. In our example, we assume you are running from this directory and use the source code a few directories up.
  • -p 8000:80 will publish port 80 on the container to port 8000 on the host, allowing you to access the HTTP server on the host interface.
  • hg-apache is the container image to run. This should correspond to what we build with docker build.

Important

The container requires that /var/hg/source contain the Mercurial source code.

Upon start, the container will attempt an install of the source in that directory. If the architecture of the host machine doesn't match that of the Docker host (e.g. when running Boot2Docker under OS X), Mercurial's Python C extensions will fail to run. Be sure to make clean your host's source tree before mounting it in the container to avoid this.

When starting the container, you should see some start-up actions (including a Mercurial install) and some output saying Apache has started:

Now if you load http://localhost:8000/ (or whatever interface Docker is using), you should see hgweb running!

For your convenience, we've created an empty repository available at /repo. Feel free to populate it with hg push.

Customizing the Server

By default, the Docker container installs a basic hgweb config and an empty dummy repository. It also uses some reasonable defaults for mod_wsgi.

Customizing the WSGI Dispatcher And Mercurial Config

By default, the Docker environment installs a custom hgweb.wsgi file (based on the example in contrib/hgweb.wsgi). The file is installed into /var/hg/htdocs/hgweb.wsgi.

A default hgweb configuration file is also installed. The hgwebconfig file from this directory is installed into /var/hg/htdocs/config.

You have a few options for customizing these files.

The simplest is to hack up hgwebconfig and entrypoint.sh in this directory and to rebuild the Docker image. This has the downside that the Mercurial working copy is modified and you may accidentally commit unwanted changes.

The next simplest is to copy this directory somewhere, make your changes, then rebuild the image. No working copy changes involved.

The preferred solution is to mount a host file into the container and overwrite the built-in defaults.

For example, say we create a custom hgweb config file in ~/hgweb. We can start the container like so to install our custom config file:

$ docker run -v ~/hgweb:/var/hg/htdocs/config ...

You can do something similar to install a custom WSGI dispatcher:

$ docker run -v ~/hgweb.wsgi:/var/hg/htdocs/hgweb.wsgi ...

Managing Repositories

Repositories are served from /var/hg/repos by default. This directory is configured as a Docker volume. This means you can mount an existing data volume container in the container so repository data is persisted across container invocations. See https://docs.docker.com/userguide/dockervolumes/ for more.

Alternatively, if you just want to perform lightweight repository manipulation, open a shell in the container:

$ docker exec -it <container> /bin/bash

Then run hg init, etc to manipulate the repositories in /var/hg/repos.

mod_wsgi Configuration Settings

mod_wsgi settings can be controlled with the following environment variables.

WSGI_PROCESSES
Number of WSGI processes to run.
WSGI_THREADS
Number of threads to run in each WSGI process
WSGI_MAX_REQUESTS
Maximum number of requests each WSGI process may serve before it is reaped.

See https://code.google.com/p/modwsgi/wiki/ConfigurationDirectives#WSGIDaemonProcess for more on these settings.

Note

The default is to use 1 thread per process. The reason is that Mercurial doesn't perform well in multi-threaded mode due to the GIL. Most people run a single thread per process in production for this reason, so that's what we default to.