Open feelotraveller opened 6 years ago
I don't know if I suggested this but you could create a personal mirror pool using
pacman-mirrors --default --interactive
Select all except the two you are having issues with. You also mention you use https only and that could cater for the responsiveness. https negotiation takes some extra time. Have you checked if the problem is the same with http?
Normally mirrors does not come and go - they stay in business for a long time - so the above approach should be better - and you could recreate the pool every 6 month or just stay with 2 or three mirrors.
If you follow the repo status site for a couple of days you should be able to locate mirrors which are well maintained and can deliver the desired responsiveness.
Personally I have only one mirror in my mirror list and I rarely, if ever change it.
Thanks for the suggestion. In the OP of the forum thread I said that that was an option but one that I did not want to take because it made my mirror pool static rather than dynamic.
I may well have overestimated the rate of change of the mirror pool since my point of comparision dates back to when the https protocol was first added and the number of https available mirrors has roughly tripled(?) :+1: since then, even though a few have disappeared. Thanks for the input, I will observe the mirror pool for a while as a point of interest.
The issue also exists in http and it is not directly one of speed. Rather the issue is with the mismatch of the speed of the inital responsiveness (the download of the 264 byte file, as you explained in the forum thread) with the download speed that occurs during an update. The two mirrors which are semi-frequently ranked the fastest (for me) are by far the slowest when actually downloading updates. (In fact they are often so slow that downloads of particular packages time out and are reinitiated by the next mirror on my list.)
Personally I prefer I bigger mirror pool because I do place a great deal of faith in governments, isp's and mirrors, generally in that order.
While I have no idea of the difficulty or work involved in adding it, and certainly do not want to be taken as in any way demanding it, I do believe that having, regardless of the solution to my current grumbles, the ability to blacklist mirrors could form a useful addition to pacman-mirrors.
certainly do not want to be taken as in any way demanding it
Indeed - I did not understand that way. I find the issue interesting from a user perspective as I am also running a mirror and the hardware is located in my local lan serving through my private internet connection.
And no I am not in business - just an ordinary user with a passion for computers :smile:
I was intrigued by your experience with one of the mirrors you mentioned so I decided I would test the philpot.de mirror. I have done so for a couple of days now and it works excellent though my location is DK.
I have no idea why you might be experiencing what you - I think - call crawl
. There might be other things involved in your issue - and I have no idea what that might be.
@fhdk suggested I make an issue here following a topic I made on the forum https://forum.manjaro.org/t/pacman-mirrors-exclude-mirror-s/47722 Feel free to amend the title if desired.
As background I use Protocols = https with no other restriction and then normally run sudo pacman-mirrors -f3 before updates. This gives me a pool of 30 plus mirrors from which generally 3 up to date mirrors are chosen.
The problem: Over a few months I have semi-regularly experienced 2 mirrors having a fast initial response times so that they get selected as fast mirrors but with subsequent slow download speeds, often to the point of timing out package downloads (10-50% occurence). (In the above thread another user mentioned experiencing similar behaviour with another mirror.)
I would like to be able to exclude these two specific mirrors while keeping my mirror pool dynamic.
The workaround suggested in that thread was to use /etc/hosts to block these mirrors. However when I reference them to 0.0.0.0 or 127.0.0.1 pacman-mirrors quite rightly throws an error 111 when trying to contact them so it is not ideal.