Issues pushing to sandbox repository - hg push either stalls or displays 404 Error Message

[Mcadams, Philip W]

Thanks Matt for pointing out the bug info.  It definitely helped me get a better understanding of our issues.

After dinking around with some options to fix this issue, I've settled upon using an IISCrypto tool on our dev server to test a fix before rolling it out on our production server.

I have a small set of users who I've chosen to test the fix.

Of the 6 users, although 4 were able to push large repos just fine 2 of them received the following type of error:

.........
searching for changes
HTTP Error: 502 (Bad Gateway)
[command returned code 255 Thu Jul 25 15:35:36 2013]

I've been Googling this error as it relates to large pushes and found posts that encourages doing incremental pushes of large repos (something I've already encouraged our developers to start doing).  Additionally on tomorrow I'm going to drill down with those users whose tests failed, which version of Mercurial they have installed on their machines.

As a side note:  A colleague on my team has communicated to me that regardless of what fix I implement, users in our environment that are located remotely have had continual issues with their network in regards to reaching our server and has felt that until we get their network upgraded users at those remote sites are going to experience issues pushing large repos.

As always, any insight that you would have on this issue would be highly appreciated.  Thank you Matt!

Philip McAdams
Systems Administrator - NVM Solutions Group Systems Engineering Apps & Infrastructure
Desk: (916) 377-6156 Cell: (916) 534-0092 Pole: FM3-1-D7




-----Original Message-----
From: Matt Mackall [mailto:mpm at selenic.com] 
Sent: Tuesday, June 11, 2013 2:43 PM
To: Mcadams, Philip W
Cc: mercurial-devel at selenic.com
Subject: Re: Issues pushing to sandbox repository - hg push either stalls or displays 404 Error Message

On Tue, 2013-06-11 at 20:51 +0000, Mcadams, Philip W wrote:
> Hello,
> 
> I'm writing regarding issues my team has encountered cloning repos 
> from sandbox to production.  Two weeks after our server migration I 
> received a JIRA ticket that the user was receiving "abort: HTTP Error
> 404: Not Found" when attempting to synchronize repos.  Another user 
> received abort: error: An existing connection was forcibly closed by 
> the remote host.
> 
> Followed advice I'd found online and I increased the IIS 7.5 
> connection timeout to 2 hours, and upped the parameters For the 
> Request Filtering Settings in IIS as well to what shown below:
> 
> [cid:image001.png at 01CE66AA.BA9937C0]
> 
> The user that initially complained of the "abort: HTTP Error 404: Not 
> Found" error appeared to have his issue resolved.  I was able to 
> reproduce his error as well and I don't get it anymore.  I did screen 
> share session with the developer and he was able to clone prod repo to 
> local pc, create sandbox repo and synchronize local pc repo with 
> sandbox repo fine: it took while as the repo is large.  Prior to 
> pushing out to our team that it's fixed I decided to try it myself.
> Although I don't get the "abort: HTTP Error 404: Not Found" but my 
> pushes just stall/hang.  I've tried pushing via cmd and with 
> Tortoisehg and they hang at different spots.  Could this be a network 
> issue?

You didn't give us any _Mercurial_ version numbers, but I suspect it's
this:

http://bz.selenic.com/show_bug.cgi?id=3905

Key indicators:

- IIS
- HTTPS
- Mercurial built against Python 2.7 (ie recent TortoiseHG)

>   A coworker informed that I need to clear the Mercurial cache.  Does 
> Mercurial have a cache that can be cleared.

There's no relevant cache nor is "clearing the cache" a thing you should ever need to do with Mercurial. Your coworker appears to be malfunctioning; you might want to try clearing their cache.

--
Mathematics is the supreme nostalgia of our time.