Socket activation for GitWeb FastCGI with systemd?

Socket activation for GitWeb FastCGI with systemd?

[Alex Ivanov]

Hi.
I want to use systemd as fastcgi spawner for gitweb + nginx. 
The traffic is low and number of users is limited + traversal bots. For that reason I've decided to use following mimimal services

gitweb.socket
[Unit]
Description=GitWeb Socket

[Socket]
ListenStream=/run/gitweb.sock
Accept=false

[Install]
WantedBy=sockets.target

gitweb.service
[Unit]
Description=GitWeb Service

[Service]
Type=simple
ExecStart=/path/to/gitweb.cgi --fcgi
StandardInput=socket

However this scheme is not resistant to simple DDOS.
E.g. traversal bots often kill the service by opening non existing path (e.g http://host/?p=repo;a=blob;f=nonexisting/path;hb=HEAD showing in browser 404 - Cannot find file) many times consecutively, which leads to
Apr 03 21:32:10 host systemd[1]: gitweb.service: Start request repeated too quickly.
Apr 03 21:32:10 host systemd[1]: gitweb.service: Failed with result 'start-limit-hit'.
Apr 03 21:32:10 host systemd[1]: Failed to start GitWeb service.
and 502 Bad Gateway in browser. I believe the reason is that gitweb.service dies on failure and if it happens too often, systemd declines to restart the service due to start limit hit.
So my question is how to correct systemd services for GitWeb to be resistant to such issue? I prefer to use single process to process all clients.
Thanks.

Re: Socket activation for GitWeb FastCGI with systemd?

[Jacob Keller]

On Tue, Apr 3, 2018 at 11:53 AM, Alex Ivanov <gnidorah@ya.ru> wrote:
> Hi.
> I want to use systemd as fastcgi spawner for gitweb + nginx.
> The traffic is low and number of users is limited + traversal bots. For that reason I've decided to use following mimimal services
>
> gitweb.socket
> [Unit]
> Description=GitWeb Socket
>
> [Socket]
> ListenStream=/run/gitweb.sock
> Accept=false
>
> [Install]
> WantedBy=sockets.target
>
> gitweb.service
> [Unit]
> Description=GitWeb Service
>
> [Service]
> Type=simple
> ExecStart=/path/to/gitweb.cgi --fcgi
> StandardInput=socket
>
> However this scheme is not resistant to simple DDOS.
> E.g. traversal bots often kill the service by opening non existing path (e.g http://host/?p=repo;a=blob;f=nonexisting/path;hb=HEAD showing in browser 404 - Cannot find file) many times consecutively, which leads to
> Apr 03 21:32:10 host systemd[1]: gitweb.service: Start request repeated too quickly.
> Apr 03 21:32:10 host systemd[1]: gitweb.service: Failed with result 'start-limit-hit'.
> Apr 03 21:32:10 host systemd[1]: Failed to start GitWeb service.
> and 502 Bad Gateway in browser. I believe the reason is that gitweb.service dies on failure and if it happens too often, systemd declines to restart the service due to start limit hit.
> So my question is how to correct systemd services for GitWeb to be resistant to such issue? I prefer to use single process to process all clients.
> Thanks.

This sounds like a systemd specific question that might get a better
answer from the systemd mailing list.

That being said, I believe if in this case gitweb is dying due to the
path not existing? You might be able to configure systemd to
understand that the particular exit code for when the path doesn't
exist is a "valid" exit, and not a failure case..

I'm not entirely understanding your goal.. you want each request to
launch the gitweb process, and when it's done you want it to exit? But
if there are multiple connections at once you want it to stay alive
until it services them all? I think the best answer is configure
systemd to understand that the exit code for when the path is invalid
will be counted as a success.

Thanks,
Jake

Re: Socket activation for GitWeb FastCGI with systemd?

[Alex Ivanov]

03.04.2018, 23:04, "Jacob Keller" <jacob.keller@gmail.com>:
> On Tue, Apr 3, 2018 at 11:53 AM, Alex Ivanov <gnidorah@ya.ru> wrote:
>>  Hi.
>>  I want to use systemd as fastcgi spawner for gitweb + nginx.
>>  The traffic is low and number of users is limited + traversal bots. For that reason I've decided to use following mimimal services
>>
>>  gitweb.socket
>>  [Unit]
>>  Description=GitWeb Socket
>>
>>  [Socket]
>>  ListenStream=/run/gitweb.sock
>>  Accept=false
>>
>>  [Install]
>>  WantedBy=sockets.target
>>
>>  gitweb.service
>>  [Unit]
>>  Description=GitWeb Service
>>
>>  [Service]
>>  Type=simple
>>  ExecStart=/path/to/gitweb.cgi --fcgi
>>  StandardInput=socket
>>
>>  However this scheme is not resistant to simple DDOS.
>>  E.g. traversal bots often kill the service by opening non existing path (e.g http://host/?p=repo;a=blob;f=nonexisting/path;hb=HEAD showing in browser 404 - Cannot find file) many times consecutively, which leads to
>>  Apr 03 21:32:10 host systemd[1]: gitweb.service: Start request repeated too quickly.
>>  Apr 03 21:32:10 host systemd[1]: gitweb.service: Failed with result 'start-limit-hit'.
>>  Apr 03 21:32:10 host systemd[1]: Failed to start GitWeb service.
>>  and 502 Bad Gateway in browser. I believe the reason is that gitweb.service dies on failure and if it happens too often, systemd declines to restart the service due to start limit hit.
>>  So my question is how to correct systemd services for GitWeb to be resistant to such issue? I prefer to use single process to process all clients.
>>  Thanks.
>
> This sounds like a systemd specific question that might get a better
> answer from the systemd mailing list.

Thanks I will try that too.

>
> That being said, I believe if in this case gitweb is dying due to the
> path not existing? You might be able to configure systemd to
> understand that the particular exit code for when the path doesn't
> exist is a "valid" exit, and not a failure case..

I will try to do that, but I'm afraid that there may be other ways to remotely abuse the service.

>
> I'm not entirely understanding your goal.. you want each request to
> launch the gitweb process, and when it's done you want it to exit? But
> if there are multiple connections at once you want it to stay alive
> until it services them all? I think the best answer is configure
> systemd to understand that the exit code for when the path is invalid
> will be counted as a success.

I want a single process for all connections too keep RAM usage at minimal. I also though it fits my case since number of users is low.

>
> Thanks,
> Jake