Mercurial checkout failure

[jer...@bodycad.com]

Hi,

I have intermittent problems with my hgweb with .cgi and apache (everything seem to work fine), I can perform a full checkout without issue with TortoiseHG under Windows machines to the web address, but sometime I got this error when doing the checkout into Jenkins:

hg pull --rev default
08:07:21 pulling from http://MyHost/MyRepos/
08:07:21 searching for changes
08:07:21 adding changesets
08:07:21 adding manifests
08:07:21 transaction abort!
08:07:21 rollback completed
08:07:21 abort: stream ended unexpectedly (got 0 bytes, expected 4)

I can start a build and end up with this and start again and it will work just fine a few minute later. The apache server access log only have 200 return code for the hgweb virtual host and the main access. Nothing into the error log. I'm a bit confuse about what can cause this, I double checked the network config of all machine imply into this and nothing seem to be wrong (ipV4, ipV6, hostname, domain) everything seem to be fine and ping work flawlessly under 0.75 ms. Anybody have a clue or a way to check what happen into those case? what is happening that close the stream (host unreachable, timeout, ...)?

Any tips is welcome here, I'm a bit at lost, I don't have anything into logs except this abort and I don't understand why.

Jenkins 2.73

Mercurial plugins 2.0

[jer...@bodycad.com]

Finally found the error, some CIFS share cache were used, now I mount all the drive with cache=none and I disabled the oplocks of the samba share, it seem to have fix the issue.