sysinit process takes 50% longer on 9.2-3 vs. 9.2-2

Having difficulties when installing the system? Your system runs slowly and requires some tweaking? You can get help here.
Post Reply

Topic author
jonesd
Master
Posts: 109
Joined: Mon Aug 09, 2021 7:59 pm
Reputation: 0
Status: Offline

sysinit process takes 50% longer on 9.2-3 vs. 9.2-2

Post by jonesd » Fri Nov 22, 2024 7:50 pm

I have a somewhat optimized boot process for the OpenVMS VM on my laptop (Virtualbox hypervisor) and after upgrading to 9.2-3, accounting records show the sysinit elapsed time went from ~15 seconds to ~23 seconds. Processor time for the process is about the same. The timings are consistent over multiple boots.

The laptop disk is an SSD. I did the upgrade to a copy of the virtual disk, could that have slowed the I/O?


hb
Master
Posts: 152
Joined: Mon May 01, 2023 12:11 pm
Reputation: 0
Status: Offline

Re: sysinit process takes 50% longer on 9.2-3 vs. 9.2-2

Post by hb » Mon Nov 25, 2024 9:29 am

jonesd wrote:
Fri Nov 22, 2024 7:50 pm
The laptop disk is an SSD. I did the upgrade to a copy of the virtual disk, could that have slowed the I/O?
The update very likely isn't it. I enabled SYSINIT messages in the boot manager. I do not see a noticeable delay. Although accounting tells me that there is a significant difference in elapsed time on V9.2-3 and V9.2-2 update 2. Unfortunately there are no timestamps in the SYSINIT messages.

From my point of view this should be report to VSI. Whether this is worth a quick fix or a fix at all, is a different question. But knowing why the accounting data is different seems to be worth it.


hb
Master
Posts: 152
Joined: Mon May 01, 2023 12:11 pm
Reputation: 0
Status: Offline

Re: sysinit process takes 50% longer on 9.2-3 vs. 9.2-2

Post by hb » Tue Nov 26, 2024 5:41 am

Hmm, ahem... I just realized that SYSINIT is not SYSINIT.
The SYSINIT process didn't change and it is done within the same elapsed time on V9.2-3 as on V9.2-2. But you don't know/see how much (elapsed) time it takes. As I already said, there are no timestamps in the SYSINIT messages.

So ACCOUNTING/TYPE=SYSINIT/FULL does not display accounting information from the SYSINIT process. It displays accounting information from the STARTUP process. Although it wasn't obvious to me, it may have been obvious to others (no, I have no other idea how to stop the Accounting utility from erasing/overwriting the screen):

Code: Select all

$ pipe accounting/since/type=sysinit/full |type sys$pipe
SYSTEM INITIALIZATION
---------------------
Username:          SYSTEM            UIC:               [SYSTEM]
Account:           <start>           Finish time:       26-NOV-2024 10:45:08.00
Process ID:        00000403          Start time:        26-NOV-2024 10:44:56.39
Owner ID:                            Elapsed time:                0 00:00:11.60
Terminal name:                       Processor time:              0 00:00:02.42
Remote node addr:                    Priority:          4
Remote node name:                    Privilege <31-00>: FFFFFFFF
Remote ID:                           Privilege <63-32>: FFFFFFFF
Remote full name:
Posix UID:         -2                Posix GID:         -2 (%XFFFFFFFE)
Queue entry:                         Final status code: 00010001
Queue name:
Job name:
Final status text: %RMS-S-NORMAL, normal successful completion
Page faults:             3868        Direct IO:               1595
Page fault reads:        2096        Buffered IO:             2845
Peak working set:       13872        Volumes mounted:            0
Peak page file:        277792        Images executed:           58
$
How do you know? It's easy. The first process in the system is the SWAPPER, the next one is SYSINIT, the third one is STARTUP. On my system this results in PIDs 401, 402 and 403.

The increase in elapsed time in TYPE=SYSINIT, that is in STARTUP is caused by starting the ACME_SERVER. This is new on V9.2-3. And this delay in STARTUP is noticeable when you watch the console output.

Post Reply