Every now and then on our Truenas and Freenas systems, we will find total NFS throughput declines to less than a megabyte/second from the usual 300-400 megabytes/second, even as half a dozen clients are getting service so slow that an ls on a tiny directory or a simple cd takes 30 seconds or more. It is only NFS that seems to be affected, and only to Linux clients. If we test with scp or a FreeBSD client, the problem is not evident. Then, after 15 minutes or half an hour, everything will be fine again for a few days or weeks. A reboot also solves the problem in deterministic time, but only temporarily.
I can find lots of advice about finding the client that is hogging the server and is therefore responsible for other users getting poor service, but that isn't the case here, since scp and Freebsd work fine. I can find worries about inadequate memory and disks with long timeouts, but again that would affect scp and FreeBSD. It is only Linux NFS that displays the problem.
I can find lots of advice about finding the client that is hogging the server and is therefore responsible for other users getting poor service, but that isn't the case here, since scp and Freebsd work fine. I can find worries about inadequate memory and disks with long timeouts, but again that would affect scp and FreeBSD. It is only Linux NFS that displays the problem.
