AIX

AIX

Connect with fellow AIX users and experts to gain knowledge, share insights, and solve problems.

 View Only
Expand all | Collapse all

AIX 7.1 transfer data problem

  • 1.  AIX 7.1 transfer data problem

    Posted Tue August 21, 2018 11:58 PM

    Originally posted by: MHRT_Tim


    Hi everyone.

    I have a problem with transfer data outgoing and incoming direction. I copied 2GB file from server A (AIX)  to another server (AIX), it took 1 minute for transfer. When I copy that file back from another server to server A, it took 15 minutes for transfer. The network does not have problem. Does anyone know how to check problem with server A because the outgoing and incoming copy speed does not equivalence.

    Thank you,

    Tim



  • 2.  Re: AIX 7.1 transfer data problem

    Posted Thu August 23, 2018 08:53 AM

    Originally posted by: savidgea


    Tim

     

    I would first double-check DNS on server A.

    Second, check storage. Are the disks local? SAN? Both the same RAID format?

    Try local writing speeds with dd command. The following will write a 2GB file in the current directory and time the operation.

     

    time dd if=/dev/zero of=./temp.out bs=1M count=1024

     

    are the times relatively similar on both? This might indicate if there's a problem with storage.



  • 3.  Re: AIX 7.1 transfer data problem

    Posted Tue August 28, 2018 02:58 AM

    Originally posted by: MHRT_Tim


    Hi Savidge,

    Server A and server B have the same subnet. The DNS does not have problem. I already test on Storage and Local disk, they copy back very slowly (both of them are quite equivalence). Server A and B's is synchronize. Speed of writing locally is as below:

     

    Thanks,



  • 4.  Re: AIX 7.1 transfer data problem

    Posted Thu August 23, 2018 01:56 PM

    Originally posted by: luverofpeanuts


     

    There are many variables that could be at play here.  In addition to checking the DNS and storage for issues, other areas to look at.....

     

    Does server A have an issue transferring back from *any* server?  Or just one in particular?

    Does server A have the same MTU setting as other servers in the transfer?  Is large send/large receive enabled on all involved?  Flow control ?  rfc1323?  

    Does the transfer take place between physical servers on real network, or is it a virtual network/virtual switch, all in the same frame?   

    Does server A own it's physical network adapter, or is it a virtual adapter served by VIOS?  Are any of the servers using micropartitioning? If so, what is the entitlement for these servers?  

    A tool like iperf can be used to to study transfer both to and from servers.   

     

     

     



  • 5.  Re: AIX 7.1 transfer data problem

    Posted Tue August 28, 2018 03:08 AM

    Originally posted by: MHRT_Tim


    Hi luverofpeanuts,

    According to your recommendation, I answer as below:

    - Just one particular. Other server transfer back is ok.

    - All the servers have the same MTU: 1500. Largesend is enabled on VIOS and LPAR

    - The server A has a virtual network (LPAR) on a VIOS (The vios has three physical port bonding with etherchannel for SEA configuration). The server B uses two 10Gbps physical network ports which is configured etherchannel with standard mode (one active, on backup adapter).

    - Server A is a LPAR and it use virtual adapter served by VIOS. The VIOS has one LPAR with CPU and Memory is high. The VIOS use three physical port with etherchannel configuration at mod 802.3ad (the switch also configured this mode) 

    - iperf command show connection refused as below: 

    app01@root:/> iperf -c 10.64.85.55
    connect failed: Connection refused

     

     



  • 6.  Re: AIX 7.1 transfer data problem

    Posted Tue August 28, 2018 11:45 AM

    Originally posted by: luverofpeanuts


     

    Hi Tim, 

     

    For iPerf  you need the 'server' side active as well.   

     

    Test for transfer from Server A to Server B

    1) On Server B (the one that transfers back slowly), run this command 'iperf -s'   (this will run the iperf in 'server' mode, and wait for client connections).   

    2) On Server A then, run 'iperf -c <ipaddress of server B>'   to run the test.  It defaults to 10 seconds.  

     

    Test for transfer from Server B to Server A

    1)  On Server A, then, do 'iperf -s',

    2)  On server B, do 'iperf -c <ipaddress of server A>'.     

     

    Summarizing what information you've provided so far. (?? note questions)

    1) MTU sizes are all 1500 -- Good

    2) Server A is LPAR with largesend enabled, uses virtual ethernet

    3) VIOS has plenty of CPU and memory, and SEA w/large send enabled, and uses 3 port etherchannel w/802.3ad mode.  (?? 10 Gbps ethernet?)

    4) Server B does not use VIOS, but uses a NIB configuration for ethernet, meaning 1 active 10 Gbps port, with 1 backup 10 Gbps port.   

    5) Server B transfers slowly, and other server(s) transfer back quickly.   

     

     

    In your original post you asked "Does anyone know how to check problem with server A",  but given the information you've provided and if I understand correctly, it seems to me like the problem is really more likely at Server B (the one that transfers back slowly).   Does that make sense to you? 

     

    More questions.... On server B, 

    1) For the active and backup 'ent' devices that are a part of the etherchannel adapter, issue this command 'lsattr -El <ent device> | grep large'   and see if large send and large receive are enabled. 

    2) Issue 'traceroute <serverA>' to make sure it's only one hop to get to server A.  

    3) Issue 'no -L | grep -e tcp_send -e tcp_recv' to see what the send and receive socket buffer sizes are. 

    4) When you transfer the file, what method do you use?   ftp?  scp?  something else?