10-18-2014
Quote:
Originally Posted by
Don Cragun
OK. So you're sending a request to a server on a socket and immediately after sending the request, you try to read a response from the server with NDELAY set in the socket options. There is no server and no network that can respond that fast to a request.
There are a couple of obvious things you could try:
- Drop the TCP_NDELAY socket option so the read() will wait for data instead of returning immediately if no data is present.
- Drop the 1st read() and start with the poll() or pollsys() to wait for data to be present before attempting the read().
I haven't tried to evaluate the arguments to
pollsys() to see if your program is waiting for data on a group of file descriptors or just waiting for data on fd #4. If it is just waiting for data on fd #4, I would start by trying #1; but if your program can continue processing if data is available on another file descriptor as well, choose option 2.
With what you have shown us there is no way for us to guess why it is taking more than 24 seconds for the server to respond to your request.
hi, thanks for reply
I have setup my own client and server on two test Solaris VMs, and my client app is able to connect to server and process data just fine. The difference is that, the test network has only one routing path and there are no intermediate routers/firewalls etc in between my client and server.
However, at my workplace, the environment is different. my production client and server VMs have multiple interfaces and so have many different routes. I am beginning to suspect it could be routing mis configuration that result in non optimal routes, or intermediate firewalls/switches causing the delay. So far, all i could think of is comparing the tcp and ip settings ( using ndd ) to that of production and see what can be fine tuned.
The client and server apps are proprietary, so I have no way to change any code.
thanks
10 More Discussions You Might Find Interesting
1. UNIX for Advanced & Expert Users
I'm a DBA so no abuse please! I've for 5 Oracle Forms processes that are spinning and am trying to find out if they're doing anything:
Running HPUX 11.11
CPU TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
9 ? 2735 oracle 241 20 24228K 16668K run 2607:29 84.92... (2 Replies)
Discussion started by: fraze
2 Replies
2. UNIX for Advanced & Expert Users
Hi,
We keep getting hanging Oracle process for our ETL. The dba's asked me to do o truss. All I see is pages and pages of the following,:cool:
pollsys(0xFFFFFFFF7FFF38C8, 1, 0xFFFFFFFF7FFF3800, 0x00000000) = 0
pollsys(0xFFFFFFFF7FFF38C8, 1, 0xFFFFFFFF7FFF3800, 0x00000000) = 0... (0 Replies)
Discussion started by: happyadm
0 Replies
3. Shell Programming and Scripting
Hi all,
while trying to debug and figure out why a lofiadm command was not working on my script, i came across a cmd called "truss"
all i know about it is that it executes the specified command and produces a trace of the system calls it performs, the signals it receives, and the machine faults... (5 Replies)
Discussion started by: wrapster
5 Replies
4. Shell Programming and Scripting
Hi
I have to get redirect the truss ouput to file.
I am doing truss -p 12121 >> output.txt
But it still displays on the screen adn output.txt is empty
Can some help how to do this?
Thanks in advance
Ammu (1 Reply)
Discussion started by: ammu
1 Replies
5. Shell Programming and Scripting
i have to gather some info about a process and redirect it to a1.txt file. For this i m using truss command
truss -po a1.txt $PID_Detail
where $PID_Detail= 1482944 3362976
--------------------------------------------------------------------------
Below the script:
#!/bin/ksh
for i... (6 Replies)
Discussion started by: ali560045
6 Replies
6. AIX
On my VIo I see the following for my disks:
$ lspath | grep hdisk6
Enabled hdisk6 fscsi0 200600a0b82193f7,4000000000000
Enabled hdisk6 fscsi0 200700a0b82193f7,4000000000000
Enabled hdisk6 fscsi2 200600a0b82193f8,4000000000000
Failed hdisk6 fscsi2 200700a0b82193f8,4000000000000
$ lspath |... (8 Replies)
Discussion started by: petervg
8 Replies
7. AIX
Hello,
I'm using AIX 5.3.12.5 and trying to understand truss output.
I'm running a job with real time of 16 minutes but only 4 minutes of CPU time. I'm trying to understand what the process is doing.
I'm getting a lot of
kread,
kpread,
kwrite,
kpwrite...
localhost:~ x$ grep... (1 Reply)
Discussion started by: Kovacs
1 Replies
8. UNIX for Dummies Questions & Answers
Hi,
I want to trace a background java program which runs in infinite loop. I have used truss command for this.
But the program terminated after some hours with below truss output:
Received signal #1, SIGHUP, in lwp_cond_wait()
/1: siginfo: SIG#0
Please let me know what... (0 Replies)
Discussion started by: Hara Prasad
0 Replies
9. UNIX for Advanced & Expert Users
Hi,
I want to trace a background java program which runs in infinite loop. I have used truss command for this.
But the program terminated after some hours with below truss output:
Received signal #1, SIGHUP, in lwp_cond_wait()
/1: siginfo: SIG#0
Please let me know what... (5 Replies)
Discussion started by: Hara Prasad
5 Replies
10. UNIX for Advanced & Expert Users
Hi,
I am trying to set ulimit for soft stack unlimited, but this is not taking effect, after tracing the ulimit -a unlimited command, the below output was generated, which i am unable to comprehend. Could any one help me with this?
prcbap1-r10prod: truss -d ulimit -s unlimited
Tue Dec 30... (2 Replies)
Discussion started by: NasirAbbasi
2 Replies
LEARN ABOUT CENTOS
rstream
RSTREAM(1) librdmacm RSTREAM(1)
NAME
rstream - streaming over RDMA ping-pong test.
SYNOPSIS
rstream [-s server_address] [-b bind_address] [-B buffer_size]
[-I iterations] [-C transfer_count]
[-S transfer_size] [-p server_port] [-T test_option]
DESCRIPTION
Uses the streaming over RDMA protocol (rsocket) to connect and exchange data between a client and server application.
OPTIONS
-s server_address
The network name or IP address of the server system listening for connections. The used name or address must route over an RDMA
device. This option must be specified by the client.
-b bind_address
The local network address to bind to.
-B buffer_size
Indicates the size of the send and receive network buffers.
-I iterations
The number of times that the specified number of messages will be exchanged between the client and server. (default 1000)
-C transfer_count
The number of messages to transfer from the client to the server and back again on each iteration. (default 1)
-S transfer_size
The size of each send transfer, in bytes. (default 1000) If 'all' is specified, rstream will run a series of tests of various
sizes.
-p server_port
The server's port number.
-T test_option
Specifies test parameters. Available options are:
s | socket - uses standard socket calls to transfer data
a | async - uses asynchronous operation (e.g. select / poll)
b | blocking - uses blocking calls
f | fork - fork server processing (forces -T s option)
n | nonblocking - uses non-blocking calls
v | verify - verifies data transfers
NOTES
Basic usage is to start rstream on a server system, then run rstream -s server_name on a client system. By default, rstream will run a
series of latency and bandwidth performance tests. Specifying a different iterations, transfer_count, or transfer_size will run a user
customized test using default values where none have been specified.
Because this test maps RDMA resources to userspace, users must ensure that they have available system resources and permissions. See the
libibverbs README file for additional details.
SEE ALSO
rdma_cm(7)
librdmacm 2011-11-16 RSTREAM(1)