A new connection will be created if (and only if) the channel arguments are different; otherwise the same connection will be reused.
On Wednesday, September 19, 2018 at 8:02:07 AM UTC-7, grpc_client wrote: > > *Language*: C++ > *gRPC version*: 1.2.5 > *Role*: gRPC client > > A quick question here. I'm observing a situation where the client attempts > to connect to a server, repeatedly printing out the following error > (SOME_IP:PORT edited out) > > [gRPC] watch_completion_error: > {"created":"@1537198702.954601066","description":"Connect > Failed","file":"src/core/ext/client_channel/subchannel.c","file_line":700,"grpc_status":14,"referenced_errors":[{"created":"@1537198702.954515638","description":"Failed > > to connect to remote host: FD > Shutdown","file":"src/core/lib/iomgr/ev_epoll_linux.c","file_line":1208,"os_error":"Timeout > > occurred","referenced_errors":[{"created":"@1537198702.954497141","description":"connect() > > timed > out","file":"src/core/lib/iomgr/tcp_client_posix.c","file_line":148}],"target_address":"ipv4:SOME_IP:PORT"}]} > > This is all fine during some init time when the server might still be > unreachable. However, this continues even when the server does become > reachable and a new grpc connection succeeds, but the initial one keeps > retrying. I can also see that the same FD is being used in all connect > attempts. > > To summarize: > > (a) there is some networking condition that prevents a successful initial > tcp connection > (b) when the condition in (a) is resolved, the initial connection still > cannot recover and continues to fail (from the logs I can see that it uses > the same FD every time for the socket) > (c) a new, 'fresh' connection is successful when (a) is a resolved, but > the initial connection is still failing > > My question is: is there a way to 'force' gRCP not to reuse the same > socket for subsequent 'connect' attempts? > > Any advice on how to mitigate the scenario above will be highly > appreciated. My current thinking is to destroy and create a new channel if > it fails repeatedly, but I was hoping for something more elegant. > > Thanks for being helpful as always! > -- You received this message because you are subscribed to the Google Groups "grpc.io" group. To unsubscribe from this group and stop receiving emails from it, send an email to grpc-io+unsubscr...@googlegroups.com. To post to this group, send email to grpc-io@googlegroups.com. Visit this group at https://groups.google.com/group/grpc-io. To view this discussion on the web visit https://groups.google.com/d/msgid/grpc-io/aa8bcdbc-25f9-456c-8517-b1f5b8d13733%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.