grunt-ssh-deploy
grunt-ssh-deploy copied to clipboard
slow upload speed
hi guys, maybe you do have an idea: I have a rather big release (150mb, gzipped with max compression). If I upload it from the shell via SCP, I get ~1.5 mb/sec. but if I upload it with grunt-ssh-deploy, I only get ~100 kb/sec which makes the deployment painfully long...
have you ever experienced something like this?
this is my config:
environments: {
options: {
local_path: "atlas-release.tar.gz",
deploy_path: "~/releases"
},
aws_test: {
options: {
host: "xxx",
username: "devenv",
port: 2222,
agent: process.env.SSH_AUTH_SOCK,
debug: true,
releases_to_keep: 1,
after_deploy: "cd ~/releases/current && tar -zxf release.tar.gz"
}
}
}
+1
+1 Пн, 14 сент. 2015 г. в 21:54, Tami Wright [email protected]:
+1
— Reply to this email directly or view it on GitHub https://github.com/dasuchin/grunt-ssh-deploy/issues/46#issuecomment-140174497 .
Until this is fixed, I'm reverting to grunt-ssh and implementing what's outlined in this blog post. If I find the time to create a patch for this package I will--and post a pull-request.
Have you tried using the zip option? The plugin itself uses scp, so I'm not sure why it would be any slower than just running scp yourself.
Yes, I did try zip and it was still extremely slow...a 12.9mb zip file wasn't uploaded after an hour. I am using a private key instead of a password if that helps at all. (A private key doesn't seem to affect how quickly *_deploy hooks execute--they are executed plenty fast...)
After moving everything over to use what I mentioned I found in the blog post (see comment before last) I had the same problems...it appears to be a sftp issue--possibly. I'm now using grunt-exec with a "scp" command to get my file on the server--and it's faster...MUCH faster.
it's definitely not about zipping content. I measured the bytes that were sent out by my PC to the network. Using grunt-ssh-deploy, only 100kb/sec are sent out to the internet, using scp directly, then it's 15x more.
If zipping the contents would have been the different, the bytes/sec should have been identical while the number of total bytes would differ.