Sindre
Sindre16mo ago

Gracefully exiting workers

I'm currently in the process of running some scheduling jobs, that I want to run quite often. Let's say every 5 sek or so. But I could also only do it once every 15 min. I'm sending a mail to new users of windmill, and nowadays people is expecting mails fast. I still run my system on docker compose. And I'm currently updating the containers quite often. docker sends SIGTERM and then SIGKILL after a grace period [1]. how does windmill workers behave when executing a job and getting a SIGTERM?
I'm never in a rush, so I do not want to kill them. my thinking is that I should configure the grace period to be the same as the TIMEOUT env var. But then I expect the workers that recive a SIGTERM to exit and not take on another job. Is this a correct understanding of how workers in windmill should work? [1] https://docs.docker.com/engine/reference/commandline/stop/
Docker Documentation
docker stop
18 Replies
rubenf
rubenf16mo ago
It will exit gracefully, wait for the current job to finish What email are you sending ? I'm thinking of adding a welcome email template :> also did you configure smtp ? Because users would get an email when they are invited to windmill, I will add an email when they login for the first time also will make it that smtp can be configured from the UI
Sindre
SindreOP16mo ago
Did not configure SMTP. I'm just checking the user table and compare with state, and sends a new email to me that I need to add them to correct a group and sends a mail to them, that they will shortly be added to a group and get access to more. I saw your comment on SSO + 50 user is EE. Would be really great if we are able to keep the license changes mostly backward compatiable. That is a huge change for me. I'm thinking we are getting EE eventually, a lot of good stuff there now, but I'm not sure it will be before we have 50 users in the system.
rubenf
rubenf16mo ago
Would 100 users work ? One slight issue that I can see with going over 50 is that now the cliff become bigger which mean you now have a choice between 0$ and a big bill versus having that choice at 50 users which is somewhat easier to digest For me there are 2 metrics that are important to track: - is it clearly an enterprise use-case, is it a medium size+ enterprise - do they have enough freedom to build conviction before paying That's how I try to decide on enterprise features
Sindre
SindreOP16mo ago
I'm more concern about new license coming up on stuff I'm already having said is included in the open source version. I see your point about 50 vs 100 and big bill. and that is def a real problem.
rubenf
rubenf16mo ago
Most tools like this will make the SSO enterprise only, that;s why I thought it was ok to be unlimited user but 50 on SSO
Sindre
SindreOP16mo ago
yeah, I see your point, but on the flip side SSO in open source was maybe the biggest selling point to choose windmill for me
rubenf
rubenf16mo ago
Also to note, I would assume that the sensitivity is not on paying or not paying, but on how much. That's something we could always negotiate We're at a point where a lot of people tell us "why would we ever pay, there is everything we need on the open-source version" which is a bit painful to hear
Sindre
SindreOP16mo ago
I think SCIM is a really good EE thing, same with Datadog/Sentry/grafana integration, if you have enought of these I think you will get EE to signup.
The problem is maybe "correlating" that with pay pr users / pr execution. That is kind of two different things. a lot of Saas is pay pr user and it hurts. If you offer us Datadog extention +++ for $499 a month, that is easierere to accecpt than yet another company that wants $ pr users, because then we do not have the same amount of control on the $. we just now it will slowly go up and up.
rubenf
rubenf16mo ago
We have SCIM 🙂 The reason for per-user is that it's correlated to usage and both the value and support you need is correlated to usage
Sindre
SindreOP16mo ago
But yes, you have a realy good open source product. monetizing on it is hard. and even harder when you like open source 🙂
rubenf
rubenf16mo ago
One big issue with only having people pay for support is that you have a huge incentive to make your product buggy and undocumented I do not want to do that, I want windmill to be 100% reliable for everyone and the docs and UX to be so clear you do not need us I don't want to limit workflow or app features except aesthetics we do not have a million levers left If we do not get profitable fast enough, we will also need to raise, which is only gonna put more pressure on monetization
Sindre
SindreOP16mo ago
haha, ppl is more lazy then you think. If you go from being a tool mostly used by developers to a tool used by non-tech people you def have a good way to help ppl with support on workflows. Think about a non tech person needing help with a script, and click a button and windmill staff comes in with multiplayer and helping him out. your workflow is extremely good.
Tiago Serafim
Tiago Serafim16mo ago
Sorry for jumping in, but please consider a SMB/Pro plan. I believe there are a lot of us that would love to pay, but aren't big enough yet.
rubenf
rubenf16mo ago
@tserafim will do, that's why right now most of the limit only hurt big enterprises but then we become a consulting company and not a tech company and the incentives are misaligned also, we do not have exclusivity on helping ppl to use windmill :>
Sindre
SindreOP16mo ago
I see your point, and you do get misaligned incentives, but I believe you will get in money. and your are not a consulting company. You get paid for helping your users with your product. nope, but you can make it the easiest way. We have a "content creation" team that only helps ppl with creation content. But all other ppl can do that also, but we know our product the best. and we are making money.
rubenf
rubenf16mo ago
Interesting, is kahoot free for everyone ? Do unis also get it for free?
Sindre
SindreOP16mo ago
so there is a lot to unpack. I can continue on PM.
rubenf
rubenf16mo ago
👍