Mercurial is slow on Macs on certain networks, except to altssh on port 443 (BB-9450)

Issue #7772 wontfix
Peter Paul Elfferich
created an issue

Hg pull, push, incoming and outgoing, etc. commands regularly take around a minute to complete on our private and paid for company repos (account: exmachina).

The repos I use are generally not large (around 10MB).

There seems to be some sort of caching involved as often, after a few successive commands the speed improves. At such a time the performance can reach a normal level where commands only take a couple of seconds to complete.

I use https to access the repos.

Strangely, when I access the same repos through the web interface I do not experience similar lag.

Here's a traceroute (I've done a few and haven't seen any noticeable spikes in those):

$ traceroute -I
traceroute: Warning: has multiple addresses; using
traceroute to (, 64 hops max, 72 byte packets
 1  my.router (  0.665 ms  0.470 ms  0.262 ms
 2 (  7.881 ms  6.724 ms  19.322 ms
 3 (  12.334 ms  10.917 ms  13.244 ms
 4 (  163.291 ms  175.255 ms  167.327 ms
 5 (  90.364 ms  91.200 ms  91.374 ms
 6 (  161.056 ms  159.259 ms  168.128 ms
 7 (  166.726 ms  161.943 ms  158.872 ms
 8 (  160.993 ms  161.075 ms  160.817 ms
 9 (  165.035 ms  162.992 ms  166.396 ms
10 (  166.690 ms  171.999 ms (  164.477 ms
11 (  164.463 ms (  169.676 ms  169.231 ms
12 (  160.125 ms  162.517 ms (  164.564 ms
13 (  168.944 ms  166.811 ms  166.683 ms
14 (  169.724 ms  160.493 ms  168.697 ms

Comments (21)

  1. Peter Paul Elfferich reporter

    Both console and SourceTree. Actually SourceTree is a bit slower still like it is in general compared to console, but that can be expected given the extra work it does.

    We've recently switched from our own little old server to bitbucket. I love the web interface and everything, but as you can imagine, along with the newer version of Mercurial and professional hosting, I was hoping for better performance, not worse.

  2. Erik van Zijst staff

    Since the data center migration we're seeing some performance issues that seem to be related to particular Internet routes.

    Are you seeing these issues just on that one location, or are you able to test it from a completely different network/ISP?

  3. Peter Paul Elfferich reporter

    That seems to be the case here indeed! I tried to pull from home a couple of times and they all performed just fine.

    Please keep me posted and feel free to let me know if I can help track those routes down.

  4. Peter Paul Elfferich reporter

    Combining all of our waiting here in the office I estimate we've lost several days of productive time to this issue by now.

    In the mean time nothing seems to be happening on your side anymore.

    Can you please at least provide us with some kind of workaround?

    Perhaps you could setup an alternate server at another location which simply tunnels all request through thus bypassing this routing issue?

  5. Jesse Yowell staff

    Thanks Peter! We may have to do some testing to see the speed results of SSH on port443 vs. 22 for Mercurial. Tell us if the problem does reoccur, and we'll look into it.

  6. Dylan Etkin

    Hi Peter,

    This issue really sounds like it is specific to your environment and that you have found a way forward that works for you.

    I am going to won't fix this and say lets run any further investigation through support. If you would like us to look into this further please email and reference this issue. We can provide a more interactive experience through the support channel and we are not seeing this as a wide-spread issue that is affecting other users.

    Thanks for understanding, cheers,


  7. Peter Paul Elfferich reporter

    Hey Dylan,

    Ok, too bad. I guess this issue is hard to pin down and fix. Will keep support in mind.

    One final note in case anybody else runs into this issue: It turns out only the Mac users in our office are affected by this, not the Windows users :S



  8. Log in to comment