Service Startup on Windows

[aarnold-sd68]

I don't think this is a bug, but an observation. In my lab I did not experience this but on physical machines it can happen. The service when set to automatic can come up before something else that is uses, not sure what this could be maybe something as simple as networking. This causes the service to register a comm error. The part that maybe a bug, is that it seems to ignore the timer and then triggers a shutdown. If I set the service to delayed start this doesn't seem to happen.

[jimklimov]

Good observation, never noticed that one. Is there any log (I'm afraid of a separate Event Log entry per what would be one syslog line) about triggering the shutdown in vain? Is there upsmon alone or also upssched?

Is there any sort of reading, maybe at driver startup, where it decides on "OB LB" (whether correctly or not) and shuts down due to that? Recently I had an outage here, and the UPS that I thought holds up for 15 minutes immediately reported that it is LB with 95% charge. I disabled upsmon for the next boot, and at 89% the UPS did really die abruptly (so it has some calibration and does know of its health, just does not propagate it into visible numbers... nor an RB status yet). So maybe that shutdown of yours is not a bug but a feature? :D

As for a comm error, any idea who failed to contact whom? I suppose localhost for upsd-upsmon link is available early. Non-networked drivers would not care about the network (but might about the USB stack; I suppose devices are detected and attached to low-level drivers like libusbK before services are attempted though).

[aarnold-sd68]

These are SNMP "connected" ups's not USB. But I'll see if I can find the even log entries for it. This is upssched and upsmon together using a timer for comm bad and on battery. This all happens in about 1 mins. The timer for upssched should be 20mins for it initiates the shutdown, so that should be enough time for everything to come up and recheck. But the server reboots before it has a chance to.

The event log looks like upsmon starts up successful, upsd reports a logon from user monusr@::1 logged into ups. then upssched reports read confirmation got [] does this a whole bunch of times then gets unable to connect to daemon and unable to start daemon, then snmp-ups starts. UPSd reports a connection. upsmon has a failure state code change from 39 to 1. upssched.exe executes command for commbad. upsmon recovered from failure state code 1 - now -1. then upsmon says user requested fsd. which comes from my upssched commbad command. upsmon auto logout and shutdown preceding. stopping upsd,stopping drivers, stopping upsmon, and edit nut. hope that helps, it never seems to run the cancel option of upssched.

[jimklimov]

I'll try to wrap my head around this, but probably not sooner than next week - now preparing for the conference.

[aarnold-sd68]

I’ve got a perfectly good work around so I’m not in any rush. I can also make it happen again 😊. From: Jim Klimov ***@***.***> Sent: Tuesday, January 27, 2026 2:57 PM To: networkupstools/nut ***@***.***> Cc: Aaron Arnold ***@***.***>; Author ***@***.***> Subject: Re: [networkupstools/nut] Service Startup on Windows (Issue #3286) CAUTION: External Message [https://avatars.githubusercontent.com/u/1636756?s=20&v=4]jimklimov left a comment (networkupstools/nut#3286)<#3286 (comment)> I'll try to wrap my head around this, but probably not sooner than next week - now preparing for the conference. — Reply to this email directly, view it on GitHub<#3286 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AQG75G5NLADQDJLV6SAZWED4I7UMPAVCNFSM6AAAAACTCGNOQKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQMBXHE2TCNZYGQ>. You are receiving this because you authored the thread.Message ID: ***@***.******@***.***>>

…

________________________________ This e-mail is privileged, confidential, subject to copyright, not intended for distribution, and may not be reproduced without the authority of the sender. Along with privileged information of the organization, this email may contain confidential personal information about students, their families or employees of NLPS. Receivers of this email are reminded that they must not forward confidential personal information to anyone who is not authorized to receive it. If you are not the intended message recipient, please contact the sender as soon as possible and do not disseminate, distribute or copy this email. Any unauthorized use or disclosure is prohibited.

[jimklimov]

For networked drivers at least, I guess you can set up a dependency (so NUT all-in-one starts after networking). This probably does not apply to all cases (e.g. standalone systems with an UPS, or those with USB/Serial links that need not wait for IP addressing to appear before becoming power-protected). See e.g. https://serverfault.com/a/228326/490516 TLDR:

sc config ServiceA depend= ServiceB/ServiceC/ServiceD/"Service Name With Spaces"

(note there is a space after the equals sign, and there is not one before it; and the depend= parameter will overwrite existing dependencies list, not append to it)

To list current dependencies: sc qc ServiceA

Digging some more into it, sc query or sc queryex (extended) alone will show you a wall of text with all services, so you can see their system names. Not sure which one would be best as a dependency for SNMP drivers, maybe Netman or iphlpsvc; maybe others for different dependencies (e.g. PlugPlay for USB?) You may have to sift through those with sc qc and build a dependency graph (probably there are tools to do so) to get an idea where you want to stick NUT in your use-case.

systemd has integration with dot IIRC to display service graphs. Googling for something similar got me a powershell script that can generate similar inputs for graphwiz: https://dkalemis.wordpress.com/2015/03/02/visualize-the-services-graph-of-your-windows-os/