You are not logged in.
I have this weird issue which is somewhat hard to pinpoint; I'm wondering if others notice similar behavior. I'm experiencing problems when writing large files over nfs (this is using vers=4.1):
# dd if=/dev/urandom of=test bs=1M count=10240
dd: error writing 'test-41': Remote I/O error
138+0 records in
137+0 records out
143654912 bytes (144 MB) copied, 26.3315 s, 5.5 MB/s
This happens consistently around the ~ 25 seconds mark. Dmesg says: nfs: server fs1 not responding, timed out. For context; here is an overview of results:
* 70254592 bytes (70 MB) copied, 24.9545 s, 2.8 MB/s
* 150994944 bytes (151 MB) copied, 27.4031 s, 5.5 MB/s
* 71303168 bytes (71 MB) copied, 24.9791 s, 2.9 MB/s
* 150994944 bytes (151 MB) copied, 26.0466 s, 5.8 MB/s
* 71303168 bytes (71 MB) copied, 25.0634 s, 2.8 MB/s
* 117440512 bytes (117 MB) copied, 23.9538 s, 4.9 MB/s
* 148897792 bytes (149 MB) copied, 26.8197 s, 5.6 MB/s
* 146800640 bytes (147 MB) copied, 25.7984 s, 5.7 MB/s
* 146800640 bytes (147 MB) copied, 25.7762 s, 5.7 MB/s
* 148897792 bytes (149 MB) copied, 26.097 s, 5.7 MB/s
* 143654912 bytes (144 MB) copied, 26.3315 s, 5.5 MB/s
* 148897792 bytes (149 MB) copied, 26.0353 s, 5.7 MB/s
The funny thing is; When I mount using vers=4.0; this time roughly halves:
dd: error writing 'test': Input/output error
150+0 records in
149+0 records out
156237824 bytes (156 MB) copied, 12.8304 s, 12.2 MB/s
This must be related to some sort of timeout somewhere; but I cannot find it. The fact that this is production; and a proper nfs-up/downgrade requires a reboot also doesn't help.
This has fixed itself... geez, thanks
Last edited by Spider.007 (2015-12-02 13:13:01)
Offline