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

I have Dropbox installed, but I don't use it.

I've got Dropbox on three Mavericks systems, with identical settings, and out of those only one is seeing this problem. It seems unlikely to be Dropbox. It could be that Dropbox show the effects of the existing problem more clearly than some other apps...

No dropbox here and never was installed. I suspect Skype indeed.

I just discovered this thread an run into the same issue for the first time today.

iMac 2012 - Mavericks

21 days uptime

No Dropbox or Skype

I didn't netstat before the reboot (I only found this thread after my rebbot :-) )

First time I am seeing this in 12 years of using OS X

Seb

Hello!

Just got to my iMac and found out I could not reach any IPv4 site but I was able to reach IPv6 sites. Then I found this thread.

Same here for me: Lots of IPv4 coonections with state FIN_WAIT_1 and lots of IPv6 dgram and stream sessions

uptime 23 days

netstat |wc -l 2990

netstat |grep FIN_WAIT |wc -l 979

and here are same values for my MBP:

uptime 2 days

netstat |wc -l 838

netstat |grep FIN_WAIT |wc -l 0

I do not use DropBox and my iMac never goes to sleep.

Hope this helps, bye

Christoph

P.S. the majority of connections is from localhost to localhost always via port 32400 - hey this is the port of the plex server version 0.9.7.28

I quit it but still 948 connections to port 32400. Someone using plex server?

I just filed bug report

15667117

Please refer to this thread and this bug number if your are calling Apple Support for this problem.

Thanks

Seb

10.9.1 update came out, I wonder if it fixes the issue. I suppose, there is no way to tell, except for waiting 3-4 weeks.

I've had this a couple of times with Mavericks

ssh: connect to host 127.0.0.1 port 22: Can't assign requested address

dtruss shows

setsockopt(0x3, 0x0, 0x3) = 0 0

connect(0x3, 0x7FEC58D02FD0, 0x10) = -1 Err#49

Under ActivityMonity kernel_task was using 6GB

But lsof and netstat don't show anything out of the ordinary.

It's as if the kernel runs out of ephemeral ports even though there appears to be no excessive use of local ports.

This also has been happening to me since 10.9, and also on 10.9.1 - on two systems. Quite frustrating. It seems to be related to network sockets/connections being consumed, but having looked at a 10.7 and a 10.9 system for sysctl, I don't see anything obvious to look at.

It just happened a few minutes ago and I rebooted before checking netstat, but that, too, is where my suspicions lie. I had about 3500 log entries for the minutes before things ground to a complete halt.

My sense is that it's not one application doing this, but, the combination of a several applications. I have a dozen IMAP sessions going, 20-30 Safari tabs, various social media clients, various cloud-storage applications, the list goes on. Those of us who have more applications consuming more networking resources are likely hitting this sooner, or, similarly, those of us with applicatoins which aren't freeing resources properly or are consuming more than others are also hitting it.

I opened a bug report with Apple as well, then found this thread - oops, should have done it the other way around.

I also noticed that if I quit applications I could regain (for a limited time) access to various things, but only rebooting really helped.

ScottM wrote:

I also noticed that if I quit applications I could regain (for a limited time) access to various things, but only rebooting really helped.

Same here.

It does not fix the issue. I have OS X 10.9.1 Build 13B42 and had the problem this morning. I've used OS X since v 1.0 and have never had this problem before. The uptime on my MBP was 20 days as of this morning. I could ping, resolve DNS, and other network activity just fine, however I was unable to pull down any HTML, either through a browser or through curl.

Later...

Richard

I have similar symptoms on a Mac Mini server, OSX 10.9.1. After about 20 days of operation my mail stopped working - all accounts at the same time - and I could not inititiate any new connections. ping and traceroute work, but telnet (for example) does not. netstat output has a large number of FIN_WAIT and CLOSE_WAIT's. And the only way out is a reboot.

I have dropbox on this machine and 2 other Mini's, and also on a MacBook pro. The Mini's are all on constantly. I don't think dropbox is at fault.

One thing that has changed recently here is the addition of 6 new remote-mounted USB drives. They are physically attached to another Mini, but this would increase the number of open sockets. (I'm a little nuts about USB storage - I have about 51 TB scattered across 22 drives. Most are USB but 6 of them are on a firewire 800 chain.)

Most of the CLOSE_WAIT's I see in the netstat output are associated with

localhost.krb524

as the foreign address.... might mean something?

Doug

I've been having this issue a lot as well. My current uptime is only 4 days.

I have suspicions about the Kerberos framework. I don't feel comfortable messing around with it, though I did see a process that implicated Kerberos and seemed to have something to do with unmounting shared drives.

I checked the authdb for rights related to com.apple.KerberosAgent and found the following:

2014-01-14 09:48:15.279 authbuddy[78694:507] Right 'com.apple.KerberosAgent' returns:

{

class = "evaluate-mechanisms";

comment = "Used to acquire Kerberos credentials.";

created = "404701328.372181";

mechanisms = (

"KerberosAgent:kerberos-dialog",

"KerberosAgent:kerberos-authenticate,privileged"

);

modified = "404701328.372181";

shared = 1;

tries = 10000;

version = 0;

}

... in case that helps anyone.

At my office I have to deal with a lot of network switching (even between ethernet and wifi), since our networking is terrible.

Here's what I regularly run on my machine:

• Dropbox
• Bitcasa
• AirVideoHD server
• Evernote
• Droplr
• VMWare Debian machine with services and NAT'd networking
• Hazel
• Time Machine backups
• Backblaze
• Messages
• Mail
• Safari
• Keyboard Maestro
• Growl
• Hazel (which keeps failing, over and over, seemingly because it can't talk with the update server)
• Network Connect (VPN software that gets disconnected all the time)
• Magic Trackpad via Bluetooth
• Some other software that check for updates

I have an early 2013 Retina MBP.

Does anyone else run any of that?

Also, note that I did kill the Kerberos-related process I found (mentioned in my last post). I also killed Dropbox, Bitcasa, Growl, Evernote, and Hazel.

I also changed my DNS servers to be 8.8.8.8 and 8.8.4.4, rather than the default (the router). I then flushed the DNS cache (`sudo dscacheutil -flushcache`) and killed (as sudo) all mDNSResponder processes.

You could try all that and see if it fixes anything for you.

I have also disabled App Nap on apps that I suspect or think will use a lot of sockets, e.g. Dropbox, VMWare, and the VPN software I use ("Network Conect"). Interestingly, you're not afforded the ability to disable App Nap for Mail or Messages.