Hi !
I'm SoftSolutions SysOp ( now back to life after 28 years in the dark ).
I've installed Synchronet 3.18 , and configured the FidoNet using Binkit.
Every time a Poll event happens, the SBBS Services consumes two vCPU fully of the
system CPU resources available ( doesn't matter the hardware host, nor how many
vCPU are allocatted , if 1 or 2 vCPUs are allocated, the system stop responding ).
When SoftSolutions receives an event from the RBT FidoNet Hub , the process happens completelly , no issues :
4/9 06:08:48p 1832 BINKP connection accepted from: 172.16.121.2 port 42326
4/9 06:08:48p 1832 BINKP BinkIT/2.39 invoked with options:
4/9 06:08:48p 1832 BINKP JSBinkP/1.123 inbound connection from 172.16.121.2:42326
4/9 06:08:48p 1832 BINKP Peer version: Mystic/1.12A46 binkp/1.0
4/9 06:08:48p 1832 BINKP Remote addresses: 4:80/1@fidonet 4:801/0@fidonet 4:80/0@fidonet
4/9 06:08:48p 1832 BINKP Inbound session for: 4:80/1@fidonet
4/9 06:08:48p 1832 BINKP Plain-text password match for 4:80/1@fidonet
4/9 06:08:49p 1832 BINKP service thread terminated (0 clients remain, 0 total, 1 served)
When a poll event happens due schedulling or manual forcing, it stops, it seems,
during the communication process :
4/10 10:13:03a BINKPOLL Running timed event: BINKPOLL
4/10 10:13:03a BINKPOLL BinkIT/2.39 invoked with options: -p
4/10 10:13:03a BINKPOLL Attempting poll for node 4:80/1@fidonet
4/10 10:13:03a BINKPOLL JSBinkP/1.123 callout to 4:80/1@fidonet started
4/10 10:13:03a BINKPOLL Connecting to 4:80/1@fidonet at fido.rbt.net.br:24554
4/10 10:13:03a BINKPOLL Will encrypt session.
4/10 10:13:03a BINKPOLL Peer version: BinkIT/2.39,JSBinkP/1.123,sbbs3.18b/Win32 binkp/1.1
4/10 10:13:03a BINKPOLL Authentication successful: non-secure
4/10 09:00:04a 0932 BINKP connection accepted from: 172.16.121.2 port 56714
4/10 09:00:04a 0932 BINKP BinkIT/2.39 invoked with options:
4/10 09:00:04a 0932 BINKP JSBinkP/1.123 inbound connection from 172.16.121.2:56714
4/10 09:00:04a 0932 BINKP Will encrypt session.
4/10 09:00:04a 0932 BINKP Peer version: BinkIT/2.39,JSBinkP/1.123,sbbs3.18b/Win32 binkp/1.1
4/10 09:00:04a 0932 BINKP Remote addresses: 4:801/10@fidonet
Calling from command line :
C:\sbbs\exec>jsexec binkit -p
JSexec v3.18b-Win32 (rev 1.217) - Execute Synchronet JavaScript Module
Compiled Sep 20 2020 18:56:02 with MSC 1927
Loading configuration files from C:\sbbs\ctrl
JavaScript-C 1.8.5 2011-03-31
JavaScript: Creating runtime: 8388608 bytes
JavaScript: Initializing context (stack: 16384 bytes)
Reading script from C:\sbbs\exec\binkit.js
C:\sbbs\exec\binkit.js compiled in 0.00 seconds
BinkIT/2.39 invoked with options: -p
Attempting poll for node 4:80/1@fidonet
JSBinkP/1.123 callout to 4:80/1@fidonet started
Connecting to 4:80/1@fidonet at fido.rbt.net.br:24554
Will encrypt session.
Peer version: BinkIT/2.39,JSBinkP/1.123,sbbs3.18b/Win32 binkp/1.1 Authentication successful: non-secure
And then the high CPU usage begins, until i recycle the service. Disabling the polling
seases the issue.
I've tried using Windows XP, Windows 2003, Windows 2008 and now Windows 2012R2 all had the same behavior, when it start a FidoNet Poll Binkit event it uses two vCPU fully.
Using the Process Explorer i see that two threads of the the sbbsNTsvcs appears to be in
some kind of loop (this explains the 2 vCPU being fully used), on the Microsoft C Runtime :
Start Address ucrtbase.DLL!_crt_at_quick_exit+0x20
both threads in this "loop" are using a full vCPU resource according the process explorer.
The only way to stop is recycling the server or kill the service.
I've made a full review on sbbsecho.ini and found nothing that is out of order, tinkit.ini is
not set (empty) on my system.
Have anyone seen this kind of behavior ? or have any tips where i can research/fix this issue ?
Regards,
AFSFigueiredo
---
þ Synchronet þ SoftSolutions -
telnet://bbs.softsolutions.net.br:2023