Hey guys we deployed 3.1 a bit ago
If you see any issues let me know I'm going to bed. We think we solved the threading issues as its been running for a while now. I'm going to be doing optimization to the client side including that caching pr.
one thing I noticed was that a hidden stack didn't appear hidden. But it's the huge stack I told you about, so maybe that's the cause of it
speed is more than nice :D
Hmm link? We kick off a background task for all operations on stacks and events
And the Ui hasn't been updated yet to use task progresses
I’d say our changes are positive so far :)
Going to keep making more changes and then release a new version for self hosters.
hey, I'm back :)
awesome!
me too
was a nice vacation
:)
My liver is glad I’m back too lol
quite difficult architectural problems at work. Leader election in dynamic, distributed systems with asynchronous unreliable networking
sucks big time.
just looking into your commits for 3.1
:(
more commits incoming over today and this week
basically we found out that signalr was a huge pain point
so we went to raw metal using persistent connection and that helped a lot
but we still noticed that issue..
the huge win is when we got rid of the redis backplane for signalr
basically we were doing double message bus messages. one that’s using our apps message bus and signalr subscribes and then signalr was using it’s own message bus which I don’t think is that well tested, we’ve always had issues with it
uh-oh
we did some thinking and we removed it because we don’t think we need it as we manage connections already and our whole app gets messages on each node so we shouldn’t need there message bus.. we did some checks and it proved to be true
their*
dang… finally the app seems to be decently fast.
keep going on and trying it and it’s staying fast.
:)
just wait until you see my next commit
and looking at the resource consumption it is a heck of a lot more efficient now too.
shaved off a request and in some cases a second request and shaved off about 2seconds on localhost
less requests and less data will be big help
yeah
kinda want to push my changes :D
lol
@niemyjski I need your help about that tsd tool :-)
hey
shoot
not because I question it, but because I want to understand what is happening there
I think you may need to add back that tsd reinstall -s (or what ever it was)
sure
so tsd is like nuget or npm but just for typescript definitions
If I understand correctly, we need that tsd reinstall to download the typings
we pull in the definitions via tsd
yes
yes. so far so good.
now, we don't have that in the developers' documentation.
so reinstall will install any missing packages because npm install tsd will only install tsd.. it won’t reinstall the definitions
we only do
npm install there
because it should happen when you do npm install
because it was a default action i had setup
:)
but then, why does it work?
if you revert that one change for tsd and do npm install (it should install the defintions or atleast that’s what I remember it doing)
because the command I think was set in the packages.config
and npm looks at that for commands to run
I think
I’m going off my memory here
yes, there is a "prepublish" command there
which does exactly that: tsd reinstall -s
and from the npm documentation, prepublish will be called on install, too.
so, isn't that enough?
yeah
it should be
users will have to do npm install when they download it
and appveyor does, too.
yep
are you not seeing this behavior
like definitions being restored
well, for me it just worked. I never had to do anything about it.
and appveyor builds work, too - my current pull request has been built even without it. so maybe we don't need it.
I guess the best thing would be to
but it's just one line - I don't really mind :-D
delete your current directory and recheck it out
and just follow the steps and if you have the definitions then it works
hehe
I did that multiple times. doesn't appveyor do a clean build every time, too?
it should yes
it’s a completely new windows vm everytime the build runs
I wanted to run something by you but I think we may have the problem solved...
?
was wondering if you ever came across something to manage / throttle requests.
so basically we are debouncing signalr notifications for x time.. but lets say a request takes xx time.. Then they start pooling up.
would be nice to just throttle a call to a single end point (like if it’s currently being made) then return that response when the promise finishes
not really. well, long time ago, in an ASP.NET HTTP Module
yeah
just wondered if we could be smarter about things
so, where does the activation start? from the server?
we want to throttle notifications?
no
it was all client side
something I saw
can you give an example?
Say we get a signalr notification saying a new event occurred..
our lists were updating it every 2.5 seconds (now moved to 10)..
well it would trigger multiple calls to refresh the dashboards and each of them could take 5 seconds….
why don't we just push the data? :-)
so now you start getting more and more and more built up pending requests because every 2.5 seconds we say refresh due to a notification
I’ve thought of that but it would be hard
been there done that.
what would be hard about it?
never really built it that way probably would take a lot to change it to do it
it would be realtime :-)
I don't know, I did both. my last multi-user desktop app had some WCF callback channel exactly for that
at first, I used CSLA to fetch the data on notification
not touching wcf with a 1000ft pole
lol
:-D
would make everything faster but not sure it’s worth it right now.
daily business. after some initial pain, one gets used to it
would change everything and cause a lot of work
yeah
it wouldn't have to change everything. it could be incorporated where most updates happen
what are the most frequent notifications?
every dashboard
but if one thing is going through it everything should
a lot of work I don’t want to take on or support right now.. maybe when we rewrite it in aurelia :)
WORD.
how far is that? haven't looked into it for some time
production ready, but is changing
i loved how Rob is putting aurelia code with angular side by side. had some good laughs
one thing I'd also like to see is some kind of indication that the UI is currently loading stuff.
this often confused me
we do have loading icons.. the grids could be better tho
any ideas
well, not putting the default text in it would be a start. The user's eyes go from "No projects were found." to "What the fuck?" to "Ah, up there, it's loading."
LOL
You know what I mean :-)
yeah
I know
I had the grid saying no data was found but eric wanted the exceptionless text
I did one project in MVC where I just put the JSON right into the view
to be honest.. I’m not happy with the whole filtering ui
I don’t like it that you don’t know what you are filtering on.
and users contact us 3 or 4 times a week saying omg I have no data what’s going on
so it loaded the JSON right from the script tag, and then got updates using the API or signalr push
and it’s cause they are filtering on something
but I guess it's 2015 and we have static UI deployment with html only.
yeah lol
would be nice to have json widgets
I’m just a nub :)
not at all, you're doing a great job.
and you know that ;)
thanks :)
still feel like a nub
lol
the loading indicators definitely need to be better.
maybe closer to the grids somehow. I don't know.
there are 2 things… 1 is initial load
which needs an indicator.
then there is we got notification that something changed… we want those to be unintrusive.
yes, right.
user should only see new data pop in.
the initial load indicator should be where the user expects the data to be
yeah, agreed.
he naturally looks there and should see some indication that something will appear there soon.
the general network working indicator and something in the grid would be good.
unobtrusive, still.
ok guys.. I think I’m going to push this and go to dinner :D
another question... how would I go about a disconnected scenario? we have a customer that doesn't want its test lab connected to the internets.
@ejsmith is here :)
@frankebersoll self host
but then we are loading scripts from cdn with no fallback
well, we can queue things up on a persisted queue...
oh I thought we were talking about the ui
lol
Self hosting in the test lab would be not very practical
@frankebersoll if you are using the .net client do client.configuration.useisolatedstorage() or client.configiguration.usefolderstorage
i thought about some USB solution.
You know… I wanted to do this in vnext
but create a relay node...
which would sit out there and just redirect requests and responses
stick in test lab, collect all the data, stick in developer pc, upload it.
yeah, that would be the alternative.
yeah, the relay thing would work.
back in the Insight product days...
we had the ability for users to send a batch of events in an email attachment.
then we had a mailbox monitor
i think i will go for the relay solution. which still means to persuade them of additional network connectivity. but that would also make continuous deployment easier.
yeah, I think that would be generally pretty useful thing to have a relay.
maybe a node process
or a TopShelf app
which would make windows service installation easier.
my customer would prefer the windows solution. what is your main user base?
never played with topshelf… but the obvious nice thing about node would be that it could be run on non-windows platforms.
which I eventually want exceptionless itself to run on non-windows as well.
obviously I am super happy whichever way you want to go.
i'll think about it. I kind of like both solutions.
good news is everything is documented now
lol
relay would be nicest but then you have to ask youself if you have that in palce they could just add the firewall rule for the domain
people are goofy… but a relay in general is nice for when the service is down.
or you could have a bunch of apps reporting to the relay and then the relay gets to batch up the events to the server.
you could even aggregate some of the events based on rules to send less stuff.
yeah
that part would be nice
what additional properties would you like to see in exceptions from JavaScript?
not sure what we are missing.
exceptionless/Exceptionless.JavaScript#8
seems like you guys have captures them pretty well.
ahh
that was a long time ago with no details… lets see if this @niemyjski guy can remember what the heck he was thinking.
he described the issue rather lean
yeah he did ;-)
hey guys
what do you have to say for yourself @niemyjski
?
@frankebersoll we don’t include the extra exception objects properties with error report
exceptionless/Exceptionless.JavaScript#9 <= this one I understand a little - what I don't understand is how those duplicates happen in the first place
so if you have a custom javascript error we don’t break it down
side note… guys check out the site right now and let me know what you think :)
can you give an example?
any faster?
@frankebersoll I’ll add comments to them in a few.
@niemyjski blazing fast
so your saying if I have a custom error that we want to snag the extra properties on it like we do in the .net client?
tremendously better.
I’m very happy. :-)
yeah
exactly eric
I was quite frankly a little embarassed by it before.
it doesn't feel sluggish any more
same I was emabarassed too
now we just need to do something about that wordpress marketing site being horribly slow.
I still got a few things I can do to speed up time to first byte on page load
and a few other perf things I have on a list
yeah
it seems to be a lot faster after I upgraded our plan
still seems like crap to me.
yeah
we need to ditch wordpress and do a github static site.
or maybe a ghost site.
yeah
would be sick
I think ghost would probably be the best.
happy medium
god dang
my hard work seems to finally be paying off
yes, it is.
I just need a vacation every other week
and progress will be made :D
@niemyjski seems to have forgotten who told him how to fix his issue. :-)
haha
who ever committed it fixed it :D
but he did help a bit
okay, I'm out for today.
man, if only that was true @niemyjski
you are getting all the credit
@frankebersoll later man!
bye!