My network connection completely died - "failed: 49 - Can't assign requested address"

Same issue here on the same machine (Mid 2012 15" Retina MBP). I tried various methods to

restore the system before giving up and rebooting.

Whats interesting is netstat took forever to print out active connections and was filled with

LAST_ACK and FIN_WAIT_1 entries. Ping worked, nslookups worked, but no new tcp/udp

sockets could be created. From the looks of it Mavericks is failing to cleanup sockets when

they are closed. I will periodically check netstat and see if this is a chronic issue or just

a transient error. If it keeps showing up I will open a bug report with Apple and see if they

can figure out what is going wrong.

I couldn't find any way to force the sockets to release from userspace.

So far my system looks clean but I will continue to monitor it to see if the issue reappears.

It appears I have the same issue, it has happened twice now. Once on my Work Mac Mini, and once on my personal Mac Mini.

So far I haven't been able to figure out what is going on. Interestingly enough IPv6 continues to function without issues, so I was able to browse Facebook/Google and various other sites, but as soon as I landed on a IPv4 only website (Apple get with the program!) it wouldn't load.

Eventually I just gave in and restarted, downside is that now I get to rebuild all of my open programs and terminals to get back to developing!

Same issue here, happened twice on the same Mac Mini.

2013-10-05: upgraded to 10.9

2013-10-30: first time happened

2013-11-18: second time

So it looks, that it takes about 20-25 days of uptime for the problem to appear on my system (it probably highly depends on the network usage, i.e. torrents, websurfing and whatnot).

I get this issue, too. Spent some time trying to see if there was a way I could manually drop connections, but seems there isn't a built-in way to do it on OSX (BSD has one; tcpdrop).

Netstat shows a lot of FIN_WAIT1, and LAST_ACK.

Interestingly, lsof doesn't show these connections, and as mentioned, ping, nslookup, and traceroute still work.

Also worth mentioning that a lot of these connections originate from a LAN IP that I no longer have. It may be related to sleeping with active connections and then changing wifi locations (This is on a retina macbook pro). Probably not applicable to Mac minis, but it might be important.

This happened to me as well. Just now. 26 days of uptime, 2012 15in MacBook Pro.

Netstat reports 714 connections, 44 of state LAST_ACK. Ping works fine, and live SSH connections stayed alive, but unable to create any new connections.

Running a Mac Pro 2010 here on 10.9 and I believe this just happened to me as well.

Could't load any sites in Safari or Chrome. Most mail connections dead. But like the OP, I could ping sites and DNS was resolving.

What's weird was my Adium connection was still up.

A logout/login didn't fix. A reboot of Airport didn't fix. Only a system reboot fixed the issue.

Honestly I can't remember when I last rebooted the system.

Anyone know a way to find out when the last reboot was? System logs only go back a week. Maybe some obscure log file that would have an entry?

EDIT:

Looking in the daily.out log, today's entry listed:

6:40 up 20 days, 12 hrs, 1 user, load averages: 0.41 0.52 0.50

So it seems to correspond with the issues starting after 20 days of uptime.

-Kevin

Message was edited by: netnothing

netnothing...

In a terminal you can get the current up time with the command uptime. Or 'last reboot' will show times of the reboots if they're still in the file. Assuming you didn't already know these.

Knew about uptime, but since I rebooted before finding this thread that wasn't going to help.

I did not know about "last reboot". That's cool....thanks!

It does confirm my uptime line I found in daily.out:

$ last reboot

reboot ~ Wed Nov 20 13:06

reboot ~ Wed Oct 30 19:40

reboot ~ Sun Oct 27 18:31

reboot ~ Sun Oct 27 18:26

reboot ~ Sun Oct 27 16:33

reboot ~ Sun Oct 27 15:51

reboot ~ Sun Oct 27 15:50

reboot ~ Tue Oct 22 22:21

reboot ~ Tue Oct 22 20:25

reboot ~ Tue Oct 22 19:59

reboot ~ Tue Oct 22 19:15

So clearly something with having a system up for ~20 days. I wonder if it's solid uptime or system uptime? My Mac Pro is on 24x7.....but my MacBook Air is usually asleep. I currently have an uptime on the Air of 15 days....so I guess in a week we'll see!

But based on previous posts, I also wonder if it has to do with overall traffic over that time as well.

-Kevin

Hmm.. I just found this thread looking for why I was seeing the same thing. This is on an iMac that's powered on 24/7.

The first signs I've been noticing are that my AppleTV has been acting odd when talking to it. Nevery every other time I'd start playback on a new movie or TV show, it would do the little pinwheel until I backed out, then restarted. Then I just saw today that the end of a file would sometimes fall back 2 or 3 levels instead of just back to the TV episode list. Then the Artwork wasn't loading in the listing. Even rebooting the aTV didn't help.

Looked at the iMac found the same situation you're all describing.

I got counts with % and saw less then the 4k counts some are seeing but counts like:

migimac(OSX) ~ {708}$ netstat -n | egrep '^(tcp|udp)' | awk '{print $NF}' | psort

0.2710% 1 SYN_SENT

0.5420% 2 CLOSING

1.3550% 5 ESTABLISHED

2.1680% 8 SYN_RCVD

10.5691% 39 FIN_WAIT_1

11.3821% 42 *.*

73.7127% 272 LAST_ACK

Total: 369

The connections to this ipv6 addr I don't recognize are odd too...

migimac(OSX) ~ {711}$ netstat -l

Active Internet connections

Proto Recv-Q Send-Q Local Address Foreign Address (state)

tcp4 0 0 migimac.miguelit.rfb 192.168.1.101.49437 ESTABLISHED

tcp6 0 820 migimac.local.daap fe80::18ec:dc40:c836:4511%en0.50694 FIN_WAIT_1

tcp6 0 220 migimac.local.daap fe80::18ec:dc40:c836:4511%en0.50691 FIN_WAIT_1

Odd that it's daap too.. It's possible my aTV is getting an IPv6 addr from my new router that's taking precedence for some reason.

I'm pretty sure uptime is just wall clock time.... i.e. now - date of last reboot = uptime.

Easy enough to test. See current uptime. Put to sleep for awhile, immediately check uptime after wake and see if it increased by the wall clock time or just the time the OS was up.

Make sure to thumbs the first post if you have the issue so we can get more eyes on this. It's a pretty serious issue if the OS doesn't release sockets.

FWIW, I also tried forcing TCP keepalive timeouts, but it didn't seem to have any affect. Does anyone know if you have to restart for those settings to take? Or if there are global minimums that are enforced for related values?

> Or if there are global minimums that are enforced for related values?

I'd look at "sysctl" settings.

It'd be great if someone would make diff of mavericks and pre-mavericks settings and post it here.

For example, keepidle is 2 hours, isn't it too much?

net.inet.tcp.keepidle: 7200000

Probably not, if it's the same as linux's net.ipv4.tcp.keepalive_time, which is also 2 hours by default on my debian system.

I had this same issue happen, on my computer at home about a week ago and on my work computer today. My uptime here at work was 31 days, but if it's that Mavericks is not releasing sockets when it should I would suspect that the length of time would be related to how much network activity is occurring that opens sockets.

Like the OP, I had no trouble pinging during this time being down. Hope this issue gets noticed and fixed soon as it's a pretty big deal since if you don't know to look in the console to troubleshoot it just looks like your internet (or network in general) has died.