On 02/23/2015 06:17 PM, Daniel Jakots wrote:

It's useful on big nodes. Due to technical limitations in how Tor handle multicore, we need to run multiple daemons on a same host so it scales better with the CPU. Even if you have AES-NI.

+1

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512

Ondrej Mikle:

...

I just filed two tickets to add support for multiple tor instances to the official deb/rpm packages.

I think this is useful for many relay operators (so they don't have to replace the init scripts themselves anymore).

I actually can't see how it can be useful except for Sybil attack. Wouldn't multiple relays per host harm network diversity?

Hi Ondrej,

I'm kinda surprised to read this from the maintainer of the RPM packages. Does that imply you will not accept patches to incorporate multi instance support in the init script shipped with RPM packages?

thanks, Nusenu

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512

Ondrej Mikle:

I still think it's a rather ugly hack that should be solved through virtualization means instead.

I'm not sure I'm following you here, could you elaborate?

But if you provide patch that works for all the supported distros (ELs and Fedoras) and doesn't negatively affect existing single-instance installations, I'd merge it.

Thanks for the info.

When it comes to the RPM packages I have to think about our past systemd migration discussion [1] that got stuck.

Should I pick that up again before writing anything that gets obsolete as soon as you migrate to systemd? (or did you give up on systemd migration?)

[1] https://lists.torproject.org/pipermail/tor-dev/2014-August/007361.html

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512

Hi everyone,

I followed the tickets about the subject and I've found the script they point to, but I didn't find any reference to a git repository, is it set up ?

On the other side the script may be useful for no so big relays too, and even for clients. One may wish to start a tor process for TBB, another for TorBirdy, another for bitcoin's core wallet, etc.

Lluís

Ondrej Mikle:

On 03/02/2015 12:22 AM, Nusenu wrote:

...

When it comes to the RPM packages I have to think about our past systemd migration discussion [1] that got stuck.

Yes I remember [2].

...

Should I pick that up again before writing anything that gets obsolete as soon as you migrate to systemd? (or did you give up on systemd migration?)

If you want to write the patch for multi-instance, don't focus on systemd. I don't plan systemd migration anytime soon since current system works and systemd migration would require systemd support on all platforms (not going to happen for EL6 any time soon).

I already had had heavily ifdef-ed packaging system because of EL5 and it was major PITA that caused some unpleasant bugs along the way.

[2] https://lists.torproject.org/pipermail/tor-dev/2014-August/007363.html

_______________________________________________ tor-relays mailing

list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512

If you want to write the patch for multi-instance, don't focus on systemd.

Thanks for the info.

I don't plan systemd migration anytime soon since current system works and systemd migration would require systemd support on all platforms (not going to happen for EL6 any time soon).

Yes, that would bring me back to my suggestion from the tor-dev thread about using upstream distro spec files and per-distro packages - but I'll leave it there and try to work something out based on the current script.