Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I'm trying to help a dev of an app I'd like to use trouble shoot a problem utilizing Corba Server on Linux. I narrowed down the problem to tnameserv taking over 3 minutes to become ready after invocation.

What exactly is tnameserv trying to do in those 3 minutes and is there anyway I can speed it up? The app failed because it tried to do 5 connection attempts with 1-second between retries; which apparently doesn't give tnameserv nearly enough time to become ready. I am using Java 6u17 on Slackware 13.0

In case it matters. The actual invocation of tnameserv is the following:

tnameserv -ORBInitialPort 23423

Upon running that command in a shell, it appears to hang until around the 3minute mark when I finally see it display "Ready".

UPDATE

I did an strace -f tnameserv -ORBInitialPort 23423 and I am seeing an eff ton of calls to gettimeofday(), clock_gettime() and futex() of which the latter is always returning '-1 ETIMEDOUT (Connection timed out). I have a feeling this is related to my problem, but I have no idea how or why.

Here is a but a small fraction of what I am seeing from strace. Can somebody replicate and/or make sense of this?

[pid 30950] futex(0x8128e14, FUTEX_WAIT_PRIVATE, 1, {0, 49903084}) = -1 ETIMEDOUT (Connection timed out)
[pid 30950] futex(0x8098a28, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329619, 995857482}) = 0
[pid 30950] gettimeofday({1260930158, 92108}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329619, 995996617}) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329619, 996088536}) = 0
[pid 30950] gettimeofday({1260930158, 92328}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_REALTIME, {1260930158, 92424295}) = 0
[pid 30950] futex(0x8128e14, FUTEX_WAIT_PRIVATE, 1, {0, 49903705}) = -1 ETIMEDOUT (Connection timed out)
[pid 30950] futex(0x8098a28, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 46761098}) = 0
[pid 30950] gettimeofday({1260930158, 143084}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 46913924}) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 47006961}) = 0
[pid 30950] gettimeofday({1260930158, 143303}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_REALTIME, {1260930158, 143398317}) = 0
[pid 30950] futex(0x8128e14, FUTEX_WAIT_PRIVATE, 1, {0, 49904683}) = -1 ETIMEDOUT (Connection timed out)
[pid 30950] futex(0x8098a28, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 97818379}) = 0
[pid 30950] gettimeofday({1260930158, 194127}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 97957235}) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 98049154}) = 0
[pid 30950] gettimeofday({1260930158, 194346}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_REALTIME, {1260930158, 194441349}) = 0
[pid 30950] futex(0x8128e14, FUTEX_WAIT_PRIVATE, 1, {0, 49904651}) = -1 ETIMEDOUT (Connection timed out)
[pid 30950] futex(0x8098a28, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 148806370}) = 0
[pid 30950] gettimeofday({1260930158, 245055}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 148947182}) = 0
[pid 30950] clock_gettime(CLOCK_MONOTONIC, {329620, 148981547}) = 0
[pid 30950] gettimeofday({1260930158, 245280}, NULL) = 0
[pid 30950] clock_gettime(CLOCK_REALTIME, {1260930158, 245374859}) = 0
[pid 30950] futex(0x8128e14, FUTEX_WAIT_PRIVATE, 1, {0, 49905141}) = -1 ETIMEDOUT (Connection timed out)
share|improve this question

2 Answers 2

I can't fix your problem directly, but experience tells me that more often than not, unexplained delays of three to five minutes are the result of waiting for DNS timeouts.

Does your configuration call for anything by hostname?

share|improve this answer
1  
Absolutely. My first response was DNS as well. Make sure your DNS is working right –  Matt Simmons Dec 16 '09 at 2:39
    
I appreciate the answer and you're right, that is something that I would have thought it would be as well but I'm 100% sure that my DNS is working because the same server I'm running 'tnameserv' on is the same server that is my firewall/squid server that I'm going through as we speak. I also did a 'dig www.google.com' successfully just to double check. –  SiegeX Dec 16 '09 at 3:29
up vote 1 down vote accepted

Turns out the problem was a firewall issue. Wireshark didn't show anything useful because the firewall was dropping a certain packet. Although I looked at my firewall logs quite a few times to make sure this wasn't the case, turns out I wasn't looking in the right place. I overlooked the fact that this 'tnameserv' was IPv6 aware (as it was binding to :::23423) and a cursory glance of my firewall script showed that I was logging IPv6 related packets to a different location than my IPv4 packets. This was not an oversight but had to be done because ip6tables does not currently support the -j ULOG target.

Long story short, allowing loopback for IPv6 fixed the issue and 'tnameserv' returns "Ready" almost instantly.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.