yet another "I have a strangeness with zypper" thread. Protocol error.... with TW zypper dup.
My zypper dup command complains a lot about: - Download error (curl) for 'https://download.opensuse.org/tumbleweed/repo/ oss/x86_64/libX11-xcb1-1.8.4-1.2.x86_64.rpm': Error code: Curl error 92 Error message: HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) A new attempt will be made... Eventually then the package will go through and the issue will repeat shortly after with some other package. This lengthens of course the procedure remarkably. One system needed the whole night to do the dup. This with TW on bare metal. My TW instance running on that machine that is KVM had a lot of curl error 55 (canceled after 5 attempts). Now my question is: does the mirror have an issue, does my modem have an issue, or do I have an issue on my local machine with some setup? By guessing I would say the mirror does have a problem with upholding the connection (maybe for server overload) or my internetprovider has a problem with the quality delivered? But you may point me to what is the cause here. Just to be sure that I can or cannot do something about it. I did have a dup of > 3200 packages, but I think this could be normal (maybe some rebuild that forced to update all....
Am 23.03.23 um 12:40 schrieb Stakanov via openSUSE Users:
My zypper dup command complains a lot about:
- Download error (curl) for 'https://download.opensuse.org/tumbleweed/repo/ oss/x86_64/libX11-xcb1-1.8.4-1.2.x86_64.rpm': Error code: Curl error 92 Error message: HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) A new attempt will be made...
Eventually then the package will go through and the issue will repeat shortly after with some other package. This lengthens of course the procedure remarkably. One system needed the whole night to do the dup.
This with TW on bare metal. My TW instance running on that machine that is KVM had a lot of curl error 55 (canceled after 5 attempts).
Now my question is: does the mirror have an issue, does my modem have an issue, or do I have an issue on my local machine with some setup?
This is a result of download.opensuse.org being overwhelmed with requests after the last TW rebuild. Nothing you can do about it.
In data giovedì 23 marzo 2023 12:43:07 CET, Benjamin Zeller ha scritto:
Am 23.03.23 um 12:40 schrieb Stakanov via openSUSE Users:
My zypper dup command complains a lot about:
- Download error (curl) for 'https://download.opensuse.org/tumbleweed/repo/ oss/x86_64/libX11-xcb1-1.8.4-1.2.x86_64.rpm': Error code: Curl error 92 Error message: HTTP/2 stream 1 was not closed cleanly: PROTOCOL_ERROR (err 1) A new attempt will be made...
Eventually then the package will go through and the issue will repeat shortly after with some other package. This lengthens of course the procedure remarkably. One system needed the whole night to do the dup.
This with TW on bare metal. My TW instance running on that machine that is KVM had a lot of curl error 55 (canceled after 5 attempts).
Now my question is: does the mirror have an issue, does my modem have an issue, or do I have an issue on my local machine with some setup?
This is a result of download.opensuse.org being overwhelmed with requests after the last TW rebuild. Nothing you can do about it.
Then I suppose that, since my Leap 15.4 PC is refusing to refresh all updates by this morning with signature fails.... on all repos is also fruit of the server going down?
On 3/23/23 08:32, Stakanov via openSUSE Users wrote:
This is a result of download.opensuse.org being overwhelmed with requests after the last TW rebuild. Nothing you can do about it. Then I suppose that, since my Leap 15.4 PC is refusing to refresh all updates by this morning with signature fails.... on all repos is also fruit of the server going down?
With Leap it will be temporary, and depends on which mirror redirect you are connecting to. Can be a few minutes to a few hours for all new packages to propagate to all mirrors and have the metatdata updated. -- David C. Rankin, J.D.,P.E.
In data giovedì 23 marzo 2023 14:49:28 CET, David C. Rankin ha scritto:
On 3/23/23 08:32, Stakanov via openSUSE Users wrote:
This is a result of download.opensuse.org being overwhelmed with requests after the last TW rebuild. Nothing you can do about it.
Then I suppose that, since my Leap 15.4 PC is refusing to refresh all updates by this morning with signature fails.... on all repos is also fruit of the server going down?
With Leap it will be temporary, and depends on which mirror redirect you are connecting to. Can be a few minutes to a few hours for all new packages to propagate to all mirrors and have the metatdata updated.
Thank you for the update. Actually is this a change of hardware? Just for curiosity, because the infrastructure up to now, at least here in Europe, did not face this kind of issues up to now.
On 2023-03-23 14:59, Stakanov via openSUSE Users wrote:
In data giovedì 23 marzo 2023 14:49:28 CET, David C. Rankin ha scritto:
On 3/23/23 08:32, Stakanov via openSUSE Users wrote:
This is a result of download.opensuse.org being overwhelmed with requests after the last TW rebuild. Nothing you can do about it.
Then I suppose that, since my Leap 15.4 PC is refusing to refresh all updates by this morning with signature fails.... on all repos is also fruit of the server going down?
With Leap it will be temporary, and depends on which mirror redirect you are connecting to. Can be a few minutes to a few hours for all new packages to propagate to all mirrors and have the metatdata updated.
Thank you for the update. Actually is this a change of hardware? Just for curiosity, because the infrastructure up to now, at least here in Europe, did not face this kind of issues up to now.
Nonono. Just that the servers are too busy. There have been this week thousands of updates on TW, and these have to be sent to the mirrors. So the server is busy both sending all those gigabytes to the mirrors and attending the requests from users. As the server (I believe) is the same for both TW and Leap, then Leap updates are also affected. This could be improved if some automated p2p method of rsync existed to propagate the updates also from mirror to mirror. It has to be invented, AFAIK. -- Cheers / Saludos, Carlos E. R. (from Elesar, using openSUSE Leap 15.4)
On 3/23/23 09:38, Carlos E. R. wrote:
Nonono.
Just that the servers are too busy.
There have been this week thousands of updates on TW, and these have to be sent to the mirrors. So the server is busy both sending all those gigabytes to the mirrors and attending the requests from users. As the server (I believe) is the same for both TW and Leap, then Leap updates are also affected.
This could be improved if some automated p2p method of rsync existed to propagate the updates also from mirror to mirror. It has to be invented, AFAIK.
May need to take a hard look at how Arch stages major gcc/glibc updates. I know since March 2009 there haven't been any similar issues that have overwhelmed the distribution servers. Rarely are daily updates more than 5-10 packages. In fact yesterday, it was 14M: Packages (14) coreutils-9.2-1 cryptsetup-2.6.1-3 debugedit-5.0-5 dvgrab-3.5-13 gpgme-1.19.0-3 gtk-update-icon-cache-1:4.10.1-2 iperf-2.1.9-1 licenses-20220125-2 python-gpgme-1.19.0-3 python-markdown-3.4.2-1 python-platformdirs-3.1.1-1 ttf-ubuntu-font-family-0.862-2 vim-9.0.1420-1 vim-runtime-9.0.1420-1 Total Download Size: 14.35 MiB Total Installed Size: 64.01 MiB Net Upgrade Size: -1.51 MiB When major changes are involved gcc/glibc, etc.. it appears from the user standpoint like they update and pre-deploy all packages that can be built compatible with the current and next gcc/glibc in the week before the actual gcc/glibc packages are released (moved from testing to core, etc..) That way, on the day the big update is released, you only get updates to OS packages directly dependent on the new version. (still can be sizable the kernel and kernel-LTS, all headers, sources, etc, but rarely exceeds 600M -- and that is a rare occurrence) Of course, if you are the type that only updates bi-weekly or monthly, then yes, you will see 1G updates each time, but for daily updates 15M - 50M are common (that's with a huge number of packages installed) Maybe opensuse could tweak the order packages are built an release to lessen the demand on the update servers? (just a thought -- that maybe what is done already) -- David C. Rankin, J.D.,P.E.
participants (5)
-
Benjamin Zeller -
Carlos E. R. -
David C. Rankin -
David C. Rankin -
Stakanov