On Thu, 28 May 2026 08:59:27 -0500 Corwin Brust <[email protected]> wrote: > > Eikes - 10m is terrible. For me things are quite snappy right now. > CGIT is entirely handled by the network of volunteer provided mirror > machines now, however this serves to distribute the DDoS from > scraping: sometimes individual mirror machines are just as bogged down > as the primary network we are trying to redirect traffic away from. > It might be worth trying again with a few minutes inbetween (or using > dig + hostfile settings to experiment with findiing a mirror that is > snappy for you). >
Yes, it is certainly an intermittent problem. Specifically, what I am seeing is a long stall, but after that initial wait, the web interface works as expected. But just now for example, as I am trying to access the cgit interface from a new location / client, I get the stall again. Fortunately, git itself is always responsive from my experience on savannah, so it's more of a minor issue with this web interface. I just wonder if things could be improved. I saw Bob's message on savannah-users last year about these volunteer-provided mirror servers. Seems to be a good system, but I have been seeing this behavior consistently over the past month or so. On IRC someone alluded to bots bogging things down, and if that is an issue, things can absolutely be alleviated with a good nftables + dynamic blacklist setup, and possibly ModSecurity + coreruleset, all on a proxy server before the traffic even reaches the web server. But I am sure the sysadmins are aware of these things, and have reasons to have things set up the way they are. Nik
