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

Oh good, at least I'm not alone in this… As per everyone else, about once a month, the machine just chokes and dies. No new net connections can be spawned, but low-level stuff like ping and mtr still run flawlessly.

FWIW, this is a quad-core 2009 Mac Pro with 12gb RAM, now running 10.9.1 but still experiencing same **** problem…

EDIT: The other log message I get lots and lots and lots of if something about sandbox denying the right to lookup com.apple.coreservices.launchservicesd.

1/14/14 9:56:37.440 PM Google Chrome Helper[66583]: Process unable to create connection because the sandbox denied the right to lookup com.apple.coreservices.launchservicesd and so this process cannot talk to launchservicesd. : LSXPCClient.cp #426 ___ZN26LSClientToServerConnection21setupServerConnectionEiPK14__CFDictionary_bl ock_invoke() q=com.apple.main-thread

Also, this machine isn't doing anything particularly strenuous. About 90% of the usage is Chrome, iTunes (streaming to Airport Express), PS3 Media Server and Mail. Mail is always hung when I try and restart, but I'm not sure that is related or not.

I don't run DropBox, btw, so that's not necessarily a culprit.

I opened bug 15824050 for this issue. I'm not sure how much Apple actually reads these forums so it would probably be a good idea if everyone opened an issue at https://bugreport.apple.com just to make sure they hear about the problem. 🙂

Later...

Richard

Yeah, I posted a bug report too.

Referenced the first bug report mentioned in this discussion and the discussion itself. Like I said, just glad to see I'm not the only one experiencing this, so there's some chance Apple will actually fix it…

This happened to me again.

Whilst not being able to make connections :-

dtruss telnet 127.0.0.1 22

socket(0x2, 0x1, 0x6) = 3 0

getuid(0x2, 0x1, 0x0) = 0 0

setuid(0x0, 0x1, 0x0) = 0 0

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

connect(0x3, 0x7F86036000E0, 0x10) = -1 Err#49 (EADDRNOTAVAIL)

netstat -na | wc -l

3218

3218 is not a lot of connections whatever they are, and telnet should be able to connect to a local address for a service that is listening.

On the presumption this is an ephemeral port problem. (see http://serverfault.com/questions/145907/does-mac-os-x-throttle-the-rate-of-socke t-creation)

I increased the number of ephemeral ports by 16384 ie

sudo sysctl -w net.inet.ip.portrange.first=32768

sudo sysctl -w net.inet.ip.portrange.hifirst=32768

Doing this brought back the ability for the machine to establish connections again, and Mail and Adium etc kicked back into life.

But after 10 minutes a different network problem manifested

Jan 14 20:24:43 mac kernel[0]: AppleBCM5701Ethernet [en0]: 8 17d LimitMinFragSize: Unable to realloc packet

Jan 14 20:24:43 mac kernel[0]: AppleBCM5701Ethernet [en0]: 0 0 transmitPacket: Failed to adjust fragment array

at this point the machine couldn't ping any address or send packets over established connections (ie ssh sessions froze), and I had to issue a reboot command from the command line (couldn't log out or switch users).

I do not see this problem as application specific, the kernel appears to be having a problem in allocating the local part of a socket, and in my case this is not because there is an excessive amount in use.

FYI, I entered bug #15825549.

FYI, I opened a bug report back on 11/20/13 - #15515590

Apple marked it as a duplicate of bug report #15264652

So they have been aware of the issue for a while now.

-Kevin

I am also experiencing this issue after 27 days uptime. I had to bounce my machine to resolve the problem.

I have the same problem, opened bug report #15832836.

Superuser question @ https://superuser.com/questions/702203/strange-network-problems-on-osx

Not sure if opening more bug reports helps but seeing the same problem on mid-2010 MBP and Mavericks 10.9.1. Disappointing to see the TCP stack so broken.

Same problem, first time it's happened to us, so I'm glad this discussion has already identified it.

- iMac 27", Late 2012; 8GB RAM; Fusion Drive; Mavericks 10.9.1, upgraded from original 10.8 (not a reformat)

- no Dropbox

- have Skype and use it sporadically, but mostly it's not running (i.e., completely Quit)

- four household users, so four users normally logged in all the time, a modest number of applications launched per user (3 to 7)

- we do not set our computer to go fully to sleep; only its display sleeps

- Time Machine backs up hourly onto a FireWire-attached drive

Amusingly, I did the same troubleshooting steps a lot of y'all did once Chrome and Safari surfing didn't work: ping, dig, traceroute, curl. All programs that wanted a TCP socket get "Can't assign requested address", including lots of things going back a few hours in my (now-gzip'ed) system.log.0. After ifconfig down/up didn't work, the only thing I could think of was to reboot, and afterward everything is working fine, and of course then I found this discussion. daily.out shows...

Local system status:

13:37 up 20 days, 13:19, 6 users, load averages: 2.82 3.10 3.29

...so if it had been up for 20d13h at 1:37pm this afternoon, and it died around 8pm, then I guess it had been up for about 20 days and 18 hours when it started failing. Like user "heavyboots", I also can see some Chrome messages about sandbox denies in the system log just before the problem; maybe this is a coincidence. I did not know that I was supposed to take a look at the netstat listing before restarting, so I don't have any counts of sockets or states or what they were connected to.

Fascinating problem. Annoying, but at least it doesn't cause me any data loss and at least it doesn't happen for many days. Apple, I hope you're listening and can find the cause of this. Chime in if you want any details we haven't provided.

-- JeffS

http://nslog.com/2013/12/14/lost_internet_connection

suggests that:-

sudo ifconfig en0 down

sudo route flush

sudo ifconfig en0 up

has brought the network back to life, after seeing this problem.

Be interested if this works for anyone.

No, this does not work for me.... In my case it is en3 because I am connected via the thunderbolt display.

Connecting the ethernet cable to the ethernet port on my mac mini itself (en0) does not solve the problem either.

Henk

gary.liddle wrote:

http://nslog.com/2013/12/14/lost_internet_connection

suggests that:-

sudo ifconfig en0 down

sudo route flush

sudo ifconfig en0 up

has brought the network back to life, after seeing this problem.

Be interested if this works for anyone.

I believe this did work for me. This is obviously very exciting, and while it's not a fix and this is still an unbelievable bug, it allows me to keep working without rebooting.

hvdm, are you sure the Ethernet port is en0? Run ifconfig on the command line and figure out which interface currently is assigned an IP and is using your router, and try this fix on that.

yes, this is exactly what I did. In case ethernet was connected to my display, en3 is used and I did "ifconfig en3 down" etc. And in case ethernat was connected to the mini directly, I used en0.

hvdm, are you sure the Ethernet port is en0? Run ifconfig on the command line and figure out which interface currently is assigned an IP and is using your router, and try this fix on that.

hvdm,

That *****; I'm bummed it didn't work. Hopefully it wasn't a fluke for me!

Out of curiosity, did you check Console.app after you tried the fix? I wonder if the errors look any different.