I've recently started using NCrunch, and am currently working on a solution with alot of tests.
I was reading about some people using an Azure VM as a Grid node and i've been trying to get it working today.
I've got NCrunch installed on the node, opened the ports and i can portqry the port on the remote VM from my dev laptop.
Trouble is after i set up the node in the Grid configuration the client seems to connect OK. Ncrunch scans the local src and starts uploading.
At discreete intervals the transfer seems to fail. It tells me "Connection Failure: Connection was closed" and hangs on maybe 80/1016 files.
In the workspace on the remote VM, some of the src has been uploaded but not all of it. I can restart it a few times but it will stop at 80/1016 again.
After a VS restart i hit 16/1016 before it stopped.
The logs on the remote VM tell me nothing. Even with high verbosity
[14:21:32.5537-Core-22] Client connection accepted from <redacted>:24351
[14:21:32.5537-Core-22] Creating server-side handler for <redacted>:24351
[14:21:32.5537-Core-22] Describing self to <redacted>:24351
I've not been able to find a local log from the grid node process.
Remco NCrunch Developer
#7738
16 Sep 2015 20:57 UTC
Hi, thanks for sharing this issue.
NCrunch pipes data over a standard TCP connection. It has a built-in timeout period using keepalive packets to try and keep the connection open when no work is being performed. The connectivity code has been pretty thoroughly tested with much larger uploads .. so why the connection is being broken in this case is somewhat of a mystery to me.
Do you have any files in your solution that are unusually large? (for example, over 100MB).
Something that may help you is that the NCrunch grid node has the ability to index and re-use files that already exist under its snapshot directory at the time the node is first initialised. Once the files are upload, the engine won't upload them again. If you can copy up the solution files to Azure manually (through FTP or some other means) then place them alongside the other files that the grid node has already uploaded into the snapshot, the node should start making use of these files the next time it is initialised. This will allow you to get past the copy step and start using the node.
NCrunch pipes data over a standard TCP connection. It has a built-in timeout period using keepalive packets to try and keep the connection open when no work is being performed. The connectivity code has been pretty thoroughly tested with much larger uploads .. so why the connection is being broken in this case is somewhat of a mystery to me.
Do you have any files in your solution that are unusually large? (for example, over 100MB).
Something that may help you is that the NCrunch grid node has the ability to index and re-use files that already exist under its snapshot directory at the time the node is first initialised. Once the files are upload, the engine won't upload them again. If you can copy up the solution files to Azure manually (through FTP or some other means) then place them alongside the other files that the grid node has already uploaded into the snapshot, the node should start making use of these files the next time it is initialised. This will allow you to get past the copy step and start using the node.
Remco wrote:Hi, thanks for sharing this issue.
NCrunch pipes data over a standard TCP connection. It has a built-in timeout period using keepalive packets to try and keep the connection open when no work is being performed. The connectivity code has been pretty thoroughly tested with much larger uploads .. so why the connection is being broken in this case is somewhat of a mystery to me.
Do you have any files in your solution that are unusually large? (for example, over 100MB).
Something that may help you is that the NCrunch grid node has the ability to index and re-use files that already exist under its snapshot directory at the time the node is first initialised. Once the files are upload, the engine won't upload them again. If you can copy up the solution files to Azure manually (through FTP or some other means) then place them alongside the other files that the grid node has already uploaded into the snapshot, the node should start making use of these files the next time it is initialised. This will allow you to get past the copy step and start using the node.
Hmm the source tree is about 3.9gb including components and blobs. No file is above 50mb though.
Hah just fixed it - had an excel data file open with changes. Seems that was blocking it somehow.
Looking forward to try this out :)
Remco NCrunch Developer
#7743
17 Sep 2015 10:27 UTC
ajohansen wrote:Remco wrote:Hi, thanks for sharing this issue.
NCrunch pipes data over a standard TCP connection. It has a built-in timeout period using keepalive packets to try and keep the connection open when no work is being performed. The connectivity code has been pretty thoroughly tested with much larger uploads .. so why the connection is being broken in this case is somewhat of a mystery to me.
Do you have any files in your solution that are unusually large? (for example, over 100MB).
Something that may help you is that the NCrunch grid node has the ability to index and re-use files that already exist under its snapshot directory at the time the node is first initialised. Once the files are upload, the engine won't upload them again. If you can copy up the solution files to Azure manually (through FTP or some other means) then place them alongside the other files that the grid node has already uploaded into the snapshot, the node should start making use of these files the next time it is initialised. This will allow you to get past the copy step and start using the node.
Hmm the source tree is about 3.9gb including components and blobs. No file is above 50mb though.
Hah just fixed it - had an excel data file open with changes. Seems that was blocking it somehow.
Looking forward to try this out :)
Interesting!!!! I'll need to review that UX. Thanks for letting me know about this :)
Post a reply
Log in to reply.