Experimental Physics and
Industrial Control System

[email protected] (Jeff Hill) · Thu, 21 Jan 1999 13:32:55 -0700

In the past long boot delays have been observed when an IOC comes up and reuses
the same IOC side dynamically assigned TCP port that it was using in the past. In this situation
some IP kernels on the server side detect out of order TCP/IP sequence numbers but will 
not disconnect until a long time out has expired. I don't know why changing the version of EPICS
would have any impact on this. Apparently this behavior varies with the version and vendor
of the IP kernel.

These problems could be reduced if vxWorks would shut down any open connections
when the user types "^x" or "reboot". Of course almost nothing can be done if it is a VME
sysreset or power down reboot. It also seems that after a reasonable time out that vxWorks could
disconnect the current FTP session and try a new dynamically assigned port, but I have not
heard that they have implemented a workaround of this nature.

Jeff

On Thursday, January 21, 1999 10:15 AM, Garrett D. Rinehart [SMTP:[email protected]] wrote:
> 
> > X-Accept-Language: en
> > MIME-Version: 1.0
> > To: "Garrett D. Rinehart" <[email protected]>
> > CC: [email protected]
> > Subject: Re: Slow booting ioc
> > Content-Transfer-Encoding: 7bit
> > 
> > "Garrett D. Rinehart" wrote:
> > > 
> > > My iocs all take much, much longer to boot under 3.13.0b11 than they did
> > > under 3.12.1. Is there a reason for this?
> > > 
> > > -- Garrett Rinehart
> > >    Intense Pulsed Neutron Source
> > >    Argonne National Laboratory
> > 
> > 
> > What does much much longer mean?
> > Which parts of the boot process takes longer?
> > 
> > Marty Kraimer
> 
> Sorry, I should have given more detail in the first place. I was just hoping someone would 
> say, "such-and-such is always the problem."
> 
> Here's the boot as seen from the shell. I've added comments (prefaced by "***") as to 
> where the problems exist.
> 
> boot device          : ei
> processor number     : 0 
> .
> .
> .
> flags (f)            : 0x0 
> startup script (s)   : st.cmd
> 
> Attaching network interface ei0... done.
> Loading... 437992 + 50684 + 27258
> 
> *** Here between the "437992" and the "+ 50684" there is a delay of one minute, sometimes 
> more. Booting from my 3.12.1 host, this line is virtually seamless.
> 
> Starting at 0x1000...
> 
> Target Name: vxTarget 
> Attaching network interface ei0... done.
> Attaching network interface lo0... done.
> Loading symbol table from
> oak:/home/oak/RINEHART/EPICS_3.13.0b11/iocBoot/iocPTS/vxWorks.sym ...done
>  
> *** Another delay of 30 seconds to one minute which used to only be a few seconds.
> 
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
>       ]]]]]]]]]]]  ]]]]     ]]]]]]]]]]       ]]              ]]]]         (R)
>  ]     ]]]]]]]]]  ]]]]]]     ]]]]]]]]       ]]               ]]]]            
>  ]]     ]]]]]]]  ]]]]]]]]     ]]]]]] ]     ]]                ]]]]            
>  ]]]     ]]]]] ]    ]]]  ]     ]]]] ]]]   ]]]]]]]]]  ]]]] ]] ]]]]  ]]   ]]]]]
>  ]]]]     ]]]  ]]    ]  ]]]     ]] ]]]]] ]]]]]]   ]] ]]]]]]] ]]]] ]]   ]]]]  
>  ]]]]]     ]  ]]]]     ]]]]]      ]]]]]]]] ]]]]   ]] ]]]]    ]]]]]]]    ]]]] 
>  ]]]]]]      ]]]]]     ]]]]]]    ]  ]]]]]  ]]]]   ]] ]]]]    ]]]]]]]]    ]]]]
>  ]]]]]]]    ]]]]]  ]    ]]]]]]  ]    ]]]   ]]]]   ]] ]]]]    ]]]] ]]]]    ]]]]
>  ]]]]]]]]  ]]]]]  ]]]    ]]]]]]]      ]     ]]]]]]]  ]]]]    ]]]]  ]]]] ]]]]]
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]       Development System
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]]
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]]       VxWorks version 5.2
>  ]]]]]]]]]]]]]]]]]]]]]]]]]]       KERNEL: WIND version 2.4
>  ]]]]]]]]]]]]]]]]]]]]]]]]]       Copyright Wind River Systems, Inc., 1984-1995
> 
>                                CPU: Motorola MVME167.  Processor #0.
>                               Memory Size: 0x800000.  BSP version 1.0.
> 
> 
> ioc_startup: tRdbTask (vxgdb) ID=  7b49dc 
> Executing startup script st.cmd ...
> # VxWorks startup file for iocPTS
> 
> ld < bin/iocCore
> *** Another 30 seconds... seems slow, but I never timed this part before.
> value = 8013836 = 0x7a480c
> 
> ld < bin/PTSsystemLib
> *** Another 10-30 seconds... seems slow, but I never timed this part before.
> value = 7879400 = 0x783ae8
> 
> # Databases
> dbLoadDatabase("dbd/PTSsystemApp.dbd")
> value = 0 = 0x0
> 
> # Record instances
> dbLoadRecords "../../PTSsystemApp/Db/ASD.database"
> value = 0 = 0x0
> 
> *** The dbLoadDatabase and dbLoadRecords seem a bit slow, but its probably just because 
> I'm usually minutes into the reboot by the time it gets here and it looks like it'll never 
> end. Assuming the later, we'll call this section "normal". 
> 
> After iocInit, I have it load some corrections that are of the form:
> < Fixes
> where Fixes is a file containing 'dbpf "record.field","value"' statements. I've noticed 
> delays of several seconds between that load command ("< Fixes") and the actual reading of 
> the commands in the file. This also never used to be the case.
> 
> On average, I'd say my ioc boot time has doubled. It used to take 2-1/2 minutes. Now it 
> rarely is to the point of loading records by that time.
> 
> 
> y the way, my ioc is an mv167, my old host is a Sparc10 with some old SunOS and the new 
> host is an Ultra-2 with Solaris 2.6. I was running EPICS_3.12.1 and now I'm running 
> 3.13.0b11.
> 
> -- Garrett
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System