14-Jun-2010

Testing the new environment
I took some time today to start setting up the new environment – but I ran into serious problems.
First, the ‘new’ machine failed to boot into SRM – hun on “f4” in the Power-On Self Test, but finally – after having removed keyboard and mouse, the system booted. I didn’t want to get to hang on IP-cluster traffic, so I booted conversational and set NISCS_USE_UDP to zero – the TCPIP database wan’t loaded but boot still hung on cluster formation. That meant I had to either disable clustering alltogether (always an option) of start the (emulated) second member. I choose for the latter – being the fastest action to perform.
Next, I started the WASD webserver – which I had built before, before the holiday 🙂 – and then did the PHP build, it took some searching but in the end, I got PHPWASD built and installed.
But testing was another story. To test it all, I would like to have PHP_INFO.PHP available – and that script – nor any other – is included in Mark Berryman’s kit. No problem, I copied the file from Diana and tried it again, using te recommended PHP.INI file.
But it fails: WASD complains it gets a non-complient CGI-header.
To find out what’s the problem, I started a WATCH session, and that showed me a severe issue; and with PHP alone in stead of PHPSHR, the same error showed up:

$ php :== $php_root:[bin]php
$ php cgi_bin:php_info.php
%DCL-W-ACTIMAGE, error activating image SSL$LIBCRYPTO_SHR32
-CLI-E-IMGNAME, image file $1$DKA0:[SYS0.SYSCOMMON.][SYSLIB]SSL$LIBCRYPTO_SHR32.EXE
-SYSTEM-F-SHRIDMISMAT, ident mismatch with shareable image
$

Asked around on the WASD mailing list and ITRC, it appeared that SSL1.4 – is NOT compatible with 1.3, and quite a lot of programs are effected. For most of them, HP has delivered patches, but for others, not yet. But Mark Berryman has updated his port of PHP and supplies objects and commandprocedures for local linkage; the same has been requested to HP.
In the mean time, Mark prepares 5.3, and when that becones available, I’ll give it a try.

Nevertheless – this means that when OpenVMS 8.4 is delivered with SSL 1.4, upgrade form 8.3 requires more than just that upgrade…

24-Feb-2010

New VMS 8.4 CD
A new edition of the 8.4 fieldtest has arrived: a backup saveset and a CD image. I burned the latter onto a rewritable disk. Nero (that I used) complained about a blocksize that didn;t match but I burned it just the same.
One thing I found on the first edition: The system, set up as an IP cluster member, fails to boot MINIMAL and waits to join the cluster. No answer – and so the system waits, and waits….
In examination, I found that what I setup in SYSBOOT, was not preserved; it might have been the setting of one SYSGEN parameter that was set to 0 where 1 is default ??? That was my second issue: I couldn’t change SYSGEN parameters in SYSBOOT….
These have been reported, and Engineering asked for a crashdump. So I made one, but since the system would not boot normally, I booted from CD and copied the dumpfile onto tape.
Next I tried to install 8.4 on a new installation of FreeAXP – but that failed drastically due to the problems on the CD…The PWS complained about bad blocksize as wel but simply did not copy the files. Three, in fact, and none seems vital (two .COM files and another).
In the mean time, Engineering could not reproduce my problem changing SYSGEN parameters during boot. That was the first thing to re-test, and I couldn’t reproduce it either. It might have been an issue that is now solved.
Now I have a system running again, the files could be retrieved from tape and uploaded to HP.

I’ve set up a new network for testing: the PWS I use has two NICs, now one connected to the ‘normal’ LAN, the other with the test network, so whatever comes along, I can access that machine.
Next, my niece’s old ACER laptop – with broken CD drive and TFT scrren, but otherwise working fine – has been setup to run Personal Alpha in the test; A bit small (just less than 1Gb of usable memory, it cannot run FreeAXP, but Personal Alpha does run, with a penalty). It’s connected to the test-LAN by wire, and to the home LAN by WiFi. On that machine, I put the .ISO file that contains the installation disk – and mentioned that as a disk container. This worked faultless – and I could install the fieldtest 8.4 kit on it’s primary disk.

A second netwrok has also been set up, but there still is something not working. No big deal, though. It’s for later use anyway.

Next stage is setting up an IP-based cluster over that test-LAN. Same LAN means that cluster-traffic over LAT will take precedence, but engineering has supplied a method to raise the priority of IP-based cluster traffic, so that limitation can now be lifted.

27-Jan-2010

Installation succceeded
The error I got on first start of FreeAXP is documented – and that proved Ok because installation of 8.4 Fieltest went smooth – abeit slow. Luckily, the workstation has a dual-core processor; like Personal Alpha, it takes one core for 100%, and it may slow down other processes, causing stalls in normal use. But it’s workable.

There is one thing to take care about when setting up a IP-based cluster: best you setup systems as unclustered, or be sure your connections over IP exist. If on the same LAN, the standard LAN communication takes precedence. And if system using IP clustering is started, it seems to expect other systems to be reachable. If not, boot will wait until another member comes along: you cannot boot an IP_based clustermember by itself….Check if the same applies when booting MINinmal…

Routing issues
As I found out: Clustering over IP won’t be feasable if the nodes are on the same LAN and SCS traffic (basicly level-2 in the ISO model) reaches ‘the other side’. LAN takes precedence over IP in those cases, and that’s why the emulator running 8.4 was clustered with Diana: Same LAN, no blocking if non_IP ttraffic. This is documented, by the way.
So I started to set up a separate network, using my new router and the old Linksys one.
It’s a hell of a job to get it running. In the end, trafic from behind the LINKSYS router (netwerk 192.168.1.0) could reach any node in the normal netwrk (192.168.0.0) but the other way around is a problem – no matter what you do. At least, PING doesn’t pass beyond the internal interface. It must have something to do with routing tables, but even that didn’t work properly. For now, there are two lines, because the Vigor router allows a seecondary address, it can be directly attached to the other network – but not using the WAN-port of the LINKSYS router. A bad solution, probably, but it works for now. Furthermore, it might be sufficient to open the UDP ports for IPCI (49152, if I’m not mistaken), but whether that allows traffic to come through remains to be seen. Also, the muticast address needs to pass the network – I know how to set that up 🙂
Running 8.4 EFT
I found a few things that seem not right – and these will first be passed to HP: The inability to boot standalone – even MINimal – when IPCI is enabled, and some problems with TCPIP commands. Both have been stored at HP (I got confirmation on these)

17-Jan-2010

Yet another old laptop
The number of old laptops to run the Alpha emulator has, again, increased. My niece switched from a Windows-base machine to MacBook – and I could get her old Acer Asipre system – the TFT screen seems dead, not any sign of output, but the virdeo hardware itself is ok, and an external screen does work – with panning, but that doen’t do any harm. It’s a Windows XP Home system but that’s fine for testing purposes. It is in thr porces of being cleaned and upgraded, and when done, I’ll put some PA disks on it. And perhaps other stuff.
VMS 8.4 testing
One of the new facilities of 8.4 is clustering the VMS systems over IP. And the system named Daphne had to be changed for that. It’s required to have it in a different network, so i had to chnage the IP configuration – to be done anyway since there had been errors from the start. It took some time, but in the end, I succeeded in doing so, but only by using the TCPIP$CONFIG procedure. Changing addresses using TCPIP ifconfig solved the problem momentarily – just in the volatile database. But the persistant data could NOT be changed. At least, I would expect it to be able to change the IP address using TCPIP SET HOST and TCPIP SET CONFIGURATION INTERFACE, but the first failed time aftre time because the host is used as an alias – correct, it is – but removing the alias wan’t possible either, for the same reason.
Moving to a standalone system was no rpoblem either, changing modparams.dat and run Autogen, but I did change the name of cluster_authorize.dat, and a direct reboot from autogen wasn’t executed. Manual reboot did work, though, and after that, I ran cluster_config_Lan to change Daphne to a clusternode, over TCPIP. The next reboot started fine, but since there isn’t a node on the other side, it hung at that point.
I’ll dig into that later.

14-Jan-2010

VMS 8.4 fieldtest
I had the systemdisk of the “new” PWS updated to 8.4, by restoring the EFT backup saveset as a file backup – not /IMAGE and though all files would have been updated, it failed to boot, BUGCHECKing during boot since SYSINIT could not be found.
Bu now I have a bootable CD – being complete – I re-applied the update; just DECWINDOWS wasn’t updated. Update suceeded – and the system (Daphne) will now boot. It’s just that the connection to the network seems to bad, I think the network cable isn’t connected to the switch. The system will be moved to another location so the connection to the LAN is right, after that I can start testing clustering over IP.
Speaking of which: The emulator on the laptop (Eden) has been setup that way, but since Diana is on the same LAN, Eden will boot into the cluster using the standard method – without TCPIP. So Daphne needs to be on a different LAN, and it needs to be a different cluster to test. So I’ll set up Daphne first, and add Eden – the 8.4 instance – into that cluster. The new router supports VLAN – as well as the new router – so interconnectr between the two LANs should be possible.
VPN test
Connection fails in authentication. The Vigor logs shows PPTP as protocol, and that wasn’t configured that way. This is still under investigation.