I’m trying to build a yocto image, but the eta is more than 2 days. The build is on the
linux-toradex-5.15.77_gitAUTOINC... do_fetch phase, and the download speed is less than 20K/s. I just ran a speed test on the same machine and same network configuration, and I got 255 MBit down.
Can you check the bandwidth on the linux-toradex git server? I’ve also opened a ticket with my internal IT to see if the issue is from our side.
I just loaded the jfrog site and the little frog icon took 14s to download the 630k png
I just did a Yocto build yesterday and didn’t notice any issues with downloads. Maybe it was just an intermittent issue? Does the slow download happen every time? Is it slow if you
git clone the repo?
This is the ip address that yocto build is connecting me to:
I tried restarting but I got the same ip address. This is also the same ipaddress that I took the screenshot from.
I just re-did the
linux-toradex-5.15.77_gitAUTOINC... do_fetch task on my side from scratch and it completed within ~15mins. With that, I don’t think there’s inherently an issue on the git server side.
Also could you answer my previous questions please. It’s not clear if this is a reoccurring issue or not and whether you have issues accessing the git server at all.
Sorry for the late reply, Monday was a holiday.
git clone git://git.toradex.com/linux-toradex.git, I get ~250 KiB, but I also hit a different server:
ip19.ip-147-135-59.us . Thru yocto I was getting one of your aws JFrog servers. I’ll investigate more on my side to see why I get 2 different servers that both serve the same thing.
Thru yocto I was getting one of your aws JFrog servers.
Possibly it’s accessing our mirrors? We have mirrors for our builds setup by default in case the original source was inaccessible at the time of fetching. For TorizonCore 6 the mirror would be our artifactory link here: Index of tdxref-torizoncore-sources-frankfurt/6
Actually now that I double-check our mirror class here: toradex-mirrors.bbclass « classes - meta-toradex-distro.git - Toradex distribution layer
I notice that these are being set as premirrors. Meaning that Yocto will actually check this location before it checks the upstream location specified by the recipe. That’s probably what is happening.
Though now the question I suppose is why your build has such a poor connection with this premirror location? As I said in my test the fetching of the kernel took ~15mins tops.
Confirmed on-premise that there’s an internal networking issue when accessing the premirrors, I’ve sent a message to our IT department. Thanks for your help @jeremias.tx !
Glad I was able to help narrow things down. Hopefully your IT team can sort things out for you.