A while back, svn updates were changed to run in parallel instead of 1 at a time. Does this (apparent) change on sourceforge's end mean that needs to be undone, at least for any sourceforge updates?
I don't have the setting to update all scripts every day enabled, so this won't affect me regardless...
My recollection behind the "go parallel" requests was that when an update failed because of a time out it took the entire sequence a very long period of time to finish. I am thinking a robust solution might stagger the requests in time and pay attention to the responses, possibly retrying.
So, for a test I added an increasing sleep at the start of CheckStatusRunnable.run (the N-th started thread waits N*100ms), and it still wasn't enough. I now got through 27 scripts before I started getting timeouts:
I think everything by Ezandora is on github?Code:Checking all SVN projects... Ezandora-Asdon-Martin-GUI-branches-Release is at HEAD (r10) mafiachit is at HEAD (r579) ccascend-cc_ascend is at HEAD (r1360) Ezandora-Helix-Fossil-branches-Release is at HEAD (r21) winterbay-mafia-wham is at HEAD (r47) stewbeef-clanfortune-branches-release is at HEAD (r32) balefull-raidlog-parser-branches-master is at HEAD (r3) mafiarecovery is at HEAD (r36) guyymafia-tricktreat is at HEAD (r274) ccascend-snapshot is at HEAD (r1360) bale-new-life is at HEAD (r70) relaywtf-skillswtf is at HEAD (r42) Ezandora-Choice-Override-branches-Release is at HEAD (r23) batman-re is at HEAD (r52) bale-counterchecker is at HEAD (r23) Ezandora-Detective-Solver-branches-Release is at HEAD (r19) clilinks is at HEAD (r6) kolmafiascripts-cfstat is at HEAD (r2) Ezandora-Genie-branches-Release is at HEAD (r44) therazekolmafia-canadv is at HEAD (r96) AllenTuring-KoL-MineVolcano-trunk is at HEAD (r61) mikebryant-kolmafia-lar-forecasting-branches-master-src is at HEAD (r54) Malurth-Auto-2-day-HCCS-branches-Release is at HEAD (r163) Ezandora-Sweet-Synthesis-branches-Release is at HEAD (r20) Ezandora-Briefcase-branches-Release is at HEAD (r80) Ezandora-Far-Future-branches-Release is at HEAD (r39) Ezandora-Guide-branches-Release is at HEAD (r516) bale-snojo not checked - exception: org.tmatesoft.svn.core.SVNException: svn: E175002: Connection has been shutdown: javax.net.ssl.SSLHandshakeException: Remote host closed connection during handshake svn: E175002: OPTIONS request failed on '/p/bale/snojo/code'
I increased the delay increase to 1000ms, and now got through 38 scripts ...
I haven't been able to get to SF at all from home for the last week. I must be on their permanent anti-DDOS list or something. Works fine from work.
Odd that they would start throttling or something without notifying the user base so that they could adjust. Also odd to begin this behavior just when they are getting a lot of migration traffic from GitHub. Some sort of causality is probably involved, but I'm not sure what can be inferred if so.
In any case, it's good to see this being discussed, thought I was the only one for a while. After verifying that DNS looked ok I started thinking maybe it was a local router / firewall problem.
I ran mafia this morning and it failed.
I saw the post above and thought I would try accessing the main page (SourceForge.net) via a web browser. I got a time out.
I then turned on a VPN for traffic coming from the browser and connected within seconds.
It sure smells like something SF is doing that is IP based. I wish someone at SF would tell people what it is.
I am having the same issue this morning needing an update to Veracity's script and getting the same errors trying to access. I did upgrade to the latest Java 10 would that be causing this?
I'm pretty sure we're somehow triggering their DDOS protection?
Also, a IP-based throttle *this* brutal can quickly block legitimate traffic from large groups behind NATs. Just imagine team of 30 people syncing the same project at the same time when they start working ...
Last edited by xKiv; 06-10-2018 at 03:54 PM.