everyone think the bot is= master
we = the girl
@Motroskin said in #30:
> The very first step to setup fail-safe solutions is to duplicate/mirror the infrastructure on independent data centers.
This is 100% right but also can become quite expensive, eating a fair share of the budget in order to prevent a rare event. Not sure it is efficient unless you can't really afford some downtime.
@rob625 said in #36:
> I would be interested to know more about the service level agreement between Lichess and OVH. It seems to have taken a long time to fix a relatively simple fault. Is there any provision for OVH to pay compensation for the downtime?
Probably, they did it when Strasbourg datacenter caught fire.
I quite enjoyed the forced break and it's good to weed out people that don't understand Lichess ' ethos , well done for getting it up and running again and all your hard work Thibault and team and don't worry , **** , it happens , sometimes xxx
This honesty and openness is awesome, and reminds me (again) why I use Lichess rather than any other platform. Modest donation on its way, just to remind you that you rock!
With due respect, any half decent sysadmin would tell that your hardware stack was too lousy, as it has a single point of failure , and it is hard to justify that by 'our budget is limited'.
As far as I can say from your blog post, you've had a single public faciing load balancer. You should have expected that sooner or later it would fail like any hardware. So you should easily have a second failover server mirroring that and ready to phase in in case of failure. It does not cost a fortune, maybe $5K to have a load balancer. Also your DC is a crappy DC. Overal a crappy French job.
Great postmortem read and outlook for the future. Appreciate the information!
@caveman4u said in #56:
> With due respect, any half decent sysadmin would tell that your hardware stack was too lousy, as it has a single point of failure , and it is hard to justify that by 'our budget is limited'.
> As far as I can say from your blog post, you've had a single public faciing load balancer. You should have expected that sooner or later it would fail like any hardware. So you should easily have a second failover server mirroring that and ready to phase in in case of failure. It does not cost a fortune, maybe $5K to have a load balancer. Also your DC is a crappy DC. Overal a crappy French job.
sound like something hit some ethnic nerve. What's the French got to do with this. Off topic, I say... All that shop talk to end with that crap ending....
It is not like our data got lost. Only fleeting games. There are many games still to be played. And the streaming is rarely lost either. Anway as it seems to be the core of lichess they managed a lifeline for it. All is safe. I would not go cultural identity bashing. (unless it satisfies some deep urges).
Actually, I had a think about this. And now I am ready to share... I pause for effect, and letting some reorient happen.
And wrapping the idea a bit. Instead of relying on central redundancy behemoth expensive double thingies. (see previous to previous shop talk but not really the last off-topic point from some person personal hell cave, to each its own demons).
yes.. i know. Instead of. ... (....) Well here is an idea of mine, I was saying I would share.
Why I might not have much sympathy for the dramatic responses might come from my selfish user case: Correspondance chess. It comes with the games. No more time of day damocles (or any time is money sentiment, or time is life or time is crappy expressing emotions). Fire your move and forget... etc..
signed the annoying correspondance lichess user. Still here... in spite of the Lobby echo chamber decorations indigestion...
Nonce