Coda File System

Re: bandwidth measure and disconnections

From: Juan Carlos Schroeder <jcsc_at_adinet.com.uy>
Date: Thu, 24 Jun 2004 12:48:54 -0300
Well, here it is what the "where" command in gdb prints after a crash when
enabling "write-back" cache in venus (a venus hang)

I forgot to say I'm using Coda Venus version 6.0.2.

----------------------------------------------

(gdb) where
#0  0x42028d69 in sigsuspend () from /lib/i686/libc.so.6
#1  0x080b7f38 in strcpy ()
#2  <signal handler called>
#3  0x4006183a in pack (a_types=0x811121c, args=0x151af4a4, _ptr=0x151af498)
    at multi2.c:403
#4  0x40061703 in pack_struct (a_types=0x811121c, args=0x4212a2d0,
    ptr=0x151af498) at multi2.c:376
#5  0x400613fa in MRPC_MakeMulti (ServerOp=60, ArgTypes=0x8111200,
HowMany=8,
    CIDList=0x81e1474, RCList=0x81e14b4, MCast=0x0, HandleResult=0xfffffffc,
    Timeout=0x0) at multi2.c:311
#6  0x080a4fa2 in strcpy ()
#7  0x08095401 in strcpy ()
#8  0x080b1577 in strcpy ()
#9  0x080b1ddc in strcpy ()
#10 0x080b6d0c in strcpy ()
#11 0x080acf25 in strcpy ()
#12 0x40083e9a in Create_Process_Part2 () at lwp.c:796
----------------------------------------------------------------------------
------

Thanks in advance,

Juan Carlos


"Juan Carlos Schroeder" <jcsc_at_adinet.com.uy> escribió en el mensaje
news:cbasp7$jlv$1_at_sea.gmane.org...
>
> I have some doubts about disconnections and the measure of bandwidth that
> venus does.
>
> My client and server are connected through a 10Mbps (local) connection.
> My server is running in a Linux VMware virtual machine inside Windows, but
> it is connected to the network as any other computer. The client is a
normal
> Linux client connected to the LAN.
>
> 1) My venus client does not really measure quite well my connection
> bandwidth.Having the full bandwidth dedicated to this, it rarely measures
> more than 30000 bytes per second (I think that's the "unit" used to show
the
> bandwidth measure that venus uses).
> It also oscilates too much, ranging from a few cents (200-300) of bytes to
a
> two or three cents of thousands (100000-200000).
> I'm reading this measures with both "codacon" executable and reading
> "venus.log" (previously configuring the debug output with "vutil d 100").
> I've also reduced bandwidth using a Windows software, and it also works
> quite bad.
>
> 2) I've had troubles to reconnect. Using "cfs strong" and then "cfs wr" it
> sometimes reconnects, but other times not. Also, some times reconnected
> after a minute or two, or even more.
> With these delays, I don't really know which command was necessary to
> reconnect.
>
> What should I do to be "connected"? What would be the sequence of
commands?
> What would be the advantages of being "connected" over being
> "write-disconnected" with "write-back" enabled ? Consistency and
> synchronization?
>
> 3) What is necessary to enable writeback mode? At first I couldn't enable
> writeback mode ("cfs wbstart"), in spite of having no "entries pending" in
> the CML. I tried to enable writeback on the volumes on the server, and
each
> time it returned me an error. I don't think writeback was enabled before
> that because I hadn't enabled this since the volume was created.
> After some time, I could, but I know why.
>
> What is necessary to enable write-back mode?
> Is it possible to be "connected" and in "write-back" mode?
>
> 4) How would Coda work with a lower "threshold" to decide to be
"connected"
> or "write-disconnected"? I plan to use low bandwidth internet links (about
> 64Kbps).
> I don't know Coda's code, but, would be hard to change this part?
> If anyone knows some utility to manage the bandwidth and latency, I'd also
> be very grateful.
>
> Thanks a lot,
>
> Juan Carlos
>
>
>
>
>
Received on 2004-06-24 11:54:24