[Adminsysters] [mur.at #9912] server disconnected

[Ralph Wozelka via RT]

and this screeny

On 20.01.22 22:23, Ralph Wozelka wrote:
> FYI, adele is maxing out its port limit hard!
> Your services cannot be reached from the outside since 30 minutes.
>
> see scrshots attached
>
> cheers,
> ralph
>
> On 20.01.22 21:38, Ralph Wozelka via RT wrote:
>> Dear adminsysters,
>>
>> around 8pm CET our core switch went into saturation again.
>>
>> this time we put a rate-limit on right onto the network port restricting to
>> 30Mbps in/out.
>> this should do the trick finally.
>> restricting traffic at our edge router was a few hops too late, obviously.
>>
>> see screenshot of traffic incoming from the core network at our edge route
>> v767-fe0-r1ko.mur.at
>>
>> let us know how adele performs on your end! :-)
>> cheers,
>> ralph
>>
>>
>>
>> On Thu Jan 20 12:48:15 2022, info at ooooo.be wrote:
>>> FIY # we did a full backup again to check the restricted rate limit we
>>> put, we expect the next backups to be incremental and much swifter.
>>>
>>> And we soon try to formulate a eeply.
>>> Thanks already for the support.
>>>
>>> XmOn Jan 19, 2022 9:40 PM, Djamil Vardag via RT <amt at mur.at> wrote:
>>>> Hey you all
>>>>> Hey ralph
>>>>> Hey mur.at
>>>>>
>>>>> We are studying to put some bandwidth limit and monitor so we can
>>>>> troubleshoot better the machine. But where to put this limit? in
>>> our
>>>>> side or in your gateway? And how much ? Like during the night we
>>> can use
>>>>> the entire bandwidth for the server upload and download..
>>>>> and during the day (8:00-22:00) Adele can use... eg. 30MB/s for
>>> upload
>>>> You can use 20 Mbit up/down, that would be ok for now. We have to
>>> discuss that
>>>> internally.
>>>>> and 10MB/s for download ? Or what do you suggest ? We need again
>>> access
>>>>> again - port 22 incoming is not enough to troubleshoot to check
>>> our borg
>>>>> backup script - for that we need for sure 22 outgoing port.
>>>> Ok, for now we have tried to implement a iptables rate limit. You
>>> should have
>>>> full TCP access to adele again, UDP is still blocked.
>>>> It would be ok for us if you use the available bandwidth as long as
>>> our
>>>> services are still working fluidly, but that isn't that easy to
>>> implement.
>>>>> It can also be our peertube instance. We disabled the webtorrent.
>>> We
>>>>> understand that peertube needs specific requirements -
>>>>> https://docs.joinpeertube.org/install-any-os?id=installation
>>>>> No adsl - you have a 100M/Bit symmetric fiber. Is that dsl? Do you
>>>> This is a 100/100 MBit fiber, no DSL, the bottleneck is the fiber
>>> and one
>>>> 100MBit switch wich will get overwhelmed if there are to many
>>> packets.
>>>>> suggest that we can run our peertube instance on your
>>> infrastructure?
>>>>> Our users are very limited and moderated. Also we don't have a lot
>>> of
>>>>> material on it. It is rather an experimental instance.
>>>> Ok, i guess this would't be a problem for our connection. Webtorrent
>>> suggests
>>>> that it uses less bandwidth on our end.
>>>>> We do understand your decision but for the moment we our server
>>> back
>>>>> running.
>>>> Sure, we can understand that and sorry that we had to be so drastic,
>>> but we
>>>> didn't have a choice as all of mur.at wasn't working well and we are
>>> loosing
>>>> our face in front of our members and financers.
>>>>> We want to install a monitoring software like netdata -
>>> promotheus.
>>>> Seems like a good idea
>>>>> Or do you have any suggestions to tackle the bandwidth problem? We
>>> have
>>>>> currently a funded project running until May so an online presence
>>> +
>>>>> visibility is importantand urgent- even while we are learning.
>>>> The problem is that we don't have any QOS on our switches enabled
>>> yet, to
>>>> tackle rate limitting there. We need to learn and implement that
>>> first.
>>>> We tried establishing a 10Mbit link to your machine, but it didn't
>>> like that.
>>>> The reasoning being that you could still use the machine, and we
>>> would have a
>>>> pseudo rate limit.
>>>> You mentioned a daily backup, maybe this is the culprit. As i said
>>> your machine
>>>> made traffic of 8TB in one week, that is just too much.
>>>> There are some websites and services of us that are also funded and
>>> we could
>>>> get a problem with renewing these services, which could cost us
>>> thousends in
>>>> the future.
>>>>
>>>>> Can you also add this email systercloud at grrlz.net to the mur.at
>>>>> mailinglist
>>>> we would suggest to put systercloud at grrlz.net as a forwarding
>>> address of donna
>>>> (donna at mur.at), who already is on all the lists.
>>>>> xm

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signal-2022-01-20-223035_001.png
Type: image/png
Size: 80512 bytes
Desc: not available
URL: <http://lists.genderchangers.org/pipermail/adminsysters/attachments/20220120/11de90b5/attachment-0001.png>