Yeah - osrm-routed uses the SO_REUSEPORT socket option (see https://lwn.net/Articles/542629/), which means you can run two copies of osrm-routed, and the operating system will round-robin incoming connections between the two processes. Handy if you want to do lossless handover from one process to another, but can cause some confusion if you forget you've already got osrm-routed running.
daniel On Mon, Apr 6, 2020 at 2:52 PM Rohit Sivakumar <[email protected]> wrote: > Please ignore the previous email. > > It appears an older osrm server on my machine had not terminated properly, > causing the nondeterminism. > > On Mon, Apr 6, 2020 at 10:34 AM Rohit Sivakumar <[email protected]> wrote: > >> Hi, >> >> I've been trying to build a profile for large vehicles on osrm. Since I >> couldn't find out-of-the-box solutions to disable all the via-way u-turns >> using just the vehicle profiles, I wrote a script last week to detect and >> add u-turn restrictions wherever possible to an .osm file passed as input. >> >> The problem is, osrm respects these restrictions sometimes, but not >> always. What's even odd is that the same Distance API call (same endpoint >> and input parameters, same profile) sometimes returns a route with u-turns >> and at other times it without the u-turns. >> >> Has anyone encountered nondeterminism in osrm before ? I can provide more >> information if this would help. >> >> Here's an osrm-frontend screenshot of one such instance where the exact >> same api-call is made to my local server, but returns two different routes: >> https://pasteboard.co/J2yZmbX.png >> > _______________________________________________ > OSRM-talk mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/osrm-talk >
_______________________________________________ OSRM-talk mailing list [email protected] https://lists.openstreetmap.org/listinfo/osrm-talk
