Language:
switch to room list switch to menu My folders
Go to page: First ... 63 64 65 66 [67] 68 69 70 71 ... Last
[#] Thu Aug 15 2013 14:14:15 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

 

Tue Aug 13 2013 11:38:30 EDT from Freakdog @ Dog Pound BBS II

 

Sun Aug 11 2013 05:11:31 PM EDT from IGnatius T Foobar @ Uncensored
Too late, I found a show stopping bug. Stay tuned!

The network spool issue that dothebart found?



hm, however much cheese I put into the trap, it won't appear anymore :(

a heisenbug...



[#] Thu Aug 15 2013 15:48:08 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Thu Aug 15 2013 14:14:15 EDT from dothebart @ Uncensored
Tue Aug 13 2013 11:38:30 EDT from Freakdog @ Dog Pound BBS II
Sun Aug 11 2013 05:11:31 PM EDT from IGnatius T Foobar @ Uncensored
Too late, I found a show stopping bug. Stay tuned!

The network spool issue that dothebart found?

hm, however much cheese I put into the trap, it won't appear anymore :(

a heisenbug...

ok, I think we caught a rat.



[#] Thu Aug 15 2013 15:55:46 EDT from vince-q @ Cascade Lodge BBS

[Reply] [ReplyQuoted] [Headers] [Print]

Ah, must be Schrödinger's Rat......

 

Thu Aug 15 2013 12:48:08 PDT from dothebart @ Uncensored

ok, I think we caught a rat.

 


[#] Thu Aug 15 2013 16:24:05 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Thu Aug 15 2013 15:55:46 EDT from vince-q @ Cascade Lodge BBS

Ah, must be Schrödinger's Rat......

Thu Aug 15 2013 12:48:08 PDT from dothebart @ Uncensored

ok, I think we caught a rat.

yes, sort of.



[#] Thu Aug 15 2013 16:33:50 EDT from Freakdog @ Dog Pound BBS II

[Reply] [ReplyQuoted] [Headers] [Print]

 

Thu Aug 15 2013 04:24:05 PM EDT from dothebart @ Uncensored
Thu Aug 15 2013 15:55:46 EDT from vince-q @ Cascade Lodge BBS

Ah, must be Schrödinger's Rat......

Thu Aug 15 2013 12:48:08 PDT from dothebart @ Uncensored

ok, I think we caught a rat.

yes, sort of.

Apparently, just another SleepyCat (see my message in Citadel Support).



[#] Thu Aug 15 2013 17:28:40 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

ok, 3c0b1d8686c302e511389e2afbbbf436af2c56e9 cures the issue.



[#] Thu Aug 15 2013 23:37:45 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

We're using the splice API? Really?

[#] Fri Aug 16 2013 01:59:57 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

if available, why not?

while sendfile may be more portable, its only working in one direction.

and... we send & receive files from the spool, so its ... PERFECT!

we also use it to cat several spool files into one. works like a charm.



[#] Fri Aug 16 2013 10:37:51 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Just questioning whether the marginal increase in speed is worth the complexity.

[#] Fri Aug 16 2013 17:26:46 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

probably not for ignet networking, but for delivering files I think yes.

in general its got the advantage, that the IO layer can handle sending & receiving of files, and the application layer doesn't need to care about that after initializing it.

and... for concatenating files, this was a speedup of > 100 times faster than before, which would fork a shell per file and call cat to concatenate them.

And, last but not least, I wanted to try and make use of this, for the shake of learning how to do it.

In the end citadel serves pretty well to build up know how & an opinion about technologies before bringing it into the game at $employer; whether its worth it, whether its to complicated to implement them in a commercial environment.

I think this served citadel pretty well so far, identifying lots of bottlenecks and removing them; finding lots of bugs and memory leaks plus increasing code quality in many places. Overcoming the strict mixing of c + html, where its pretty hard to find people willing to contribute good (or even able to do so) at both. And, a very powerfull client application stack, which you probably can abuse to load test other environments. And all this scalable enough that even people with small hardware are pretty fond of running citadel.

Plus, it serves as my mail server, where some more guinea pigs tested mailinglist functionality long before it was usefull enough for uncensored and citadel support.

so, all in all, lots of fun.



[#] Sat Aug 17 2013 12:49:27 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Fun perhaps, but if it is introducing the kind of bugs that Vince and Freakdog are experiencing, it isn't worth it ...

How many virtual Citadel nodes are you testing the networking changes with?
Unless it's at least 3 it's not enough ...

[#] Sat Aug 17 2013 13:07:50 EDT from vince-q @ Cascade Lodge BBS

[Reply] [ReplyQuoted] [Headers] [Print]

 

Sat Aug 17 2013 09:49:27 PDT from IGnatius T Foobar @ Uncensored
Fun perhaps, but if it is introducing the kind of bugs that Vince and Freakdog are experiencing, it isn't worth it ...

How many virtual Citadel nodes are you testing the networking changes with?
Unless it's at least 3 it's not enough ...

I have indicated to freakdog that the problem we (his node and mine) are having is because *his* Citadel is running inside of "virtual-box" and I contend *that* is the cause of the problem(s). Freakdog disagrees.

--Vince



[#] Sat Aug 17 2013 14:14:53 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

 

Sat Aug 17 2013 13:07:50 EDT from vince-q @ Cascade Lodge BBS

 

Sat Aug 17 2013 09:49:27 PDT from IGnatius T Foobar @ Uncensored
Fun perhaps, but if it is introducing the kind of bugs that Vince and Freakdog are experiencing, it isn't worth it ...

How many virtual Citadel nodes are you testing the networking changes with?
Unless it's at least 3 it's not enough ...

I have indicated to freakdog that the problem we (his node and mine) are having is because *his* Citadel is running inside of "virtual-box" and I contend *that* is the cause of the problem(s). Freakdog disagrees.

--Vince

we're not in the church here - evidence counts. maybe we need some more logging about filenames in question here?

there have been race conditions in the citadel networking code until 8.00 for over a decade, which caused crashes now and then - which i've fixed.

 

as you can see in cit sup, its working across 3 hops - potzblitz - outgesourced - uncensored, and it seems as if db2 cascade logde also worked.

 

 



[#] Sat Aug 17 2013 15:19:34 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Sat Aug 17 2013 14:14:53 EDT from dothebart @ Uncensored

we're not in the church here - evidence counts. maybe we need some more logging about filenames in question here?

998f7ba061f498ff16d542409ffe8926671d92ea & 8932444317c238704ce25ca3ddfcc3cc65f63dfa

adds these log outputs so you should see where files come from, and where they're going.



[#] Sat Aug 17 2013 15:27:59 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

btw, the splice code came with 8.1x



[#] Sat Aug 17 2013 17:12:47 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I have indicated to freakdog that the problem we (his node and mine)
are having is because *his* Citadel is running inside of
"virtual-box" and I contend *that* is the cause of the problem(s).
Freakdog disagrees.

I suppose the only way to be sure of that would be to try it both ways, but ... I can't think of any possible way that a userspace program like Citadel Server would care whether a block device two layers down is physical or virtual.
Anything's possible, I guess...

[#] Sat Aug 17 2013 22:57:37 EDT from vince-q @ Cascade Lodge BBS

[Reply] [ReplyQuoted] [Headers] [Print]

 

Sat Aug 17 2013 14:12:47 PDT from IGnatius T Foobar @ Uncensored
I have indicated to freakdog that the problem we (his node and mine)
are having is because *his* Citadel is running inside of
"virtual-box" and I contend *that* is the cause of the problem(s).
Freakdog disagrees.

I suppose the only way to be sure of that would be to try it both ways, but ... I can't think of any possible way that a userspace program like Citadel Server would care whether a block device two layers down is physical or virtual.
Anything's possible, I guess...

In my opinion, and please remember that I am returning to this stuff after a hiatus of just over 12 years, it is not an issue of network layers failing to communicate.

In this specific instance I believe it is a shortcoming in the virtual-box software itself.

I am seeing time-out messages in my Aide> room during sessions initiated by CASCADE aimed at DOGPOUND2.

When DOGPOUND2 "calls" here, apparently everything goes smoothly.

Now, on to "virtual box." I've seen that thing lose "communication" to the mouse, to USB flashdrives, to USB keyboards, to the NIC and hence to the outside world - at random, and with sufficient frequency that I refuse to use it - period. The level of user frustration far exceeds what I find to be acceptable.

I am sorry if this message offends. That is absolutely not my intention in any way.

Perhaps I am completely "off base" in my "indictment" of virtual-box. However from what I've seen of it, I am several orders of magnitude less than impressed.

--Vince (K2NE)



[#] Mon Aug 19 2013 00:04:38 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Ok, that makes more sense -- you are suspecting the VirtualBox software specifically, not virtualization in general. Uncensored has been running virtualized for about six years now, originally on Virtuozzo and then briefly on VMware but mostly (and currently) on KVM.

I actually suspect that the "performance optimizations" made to that layer will probably have something to do with it, but I will leave that to dothebart to figure out.

[#] Mon Aug 19 2013 02:13:14 EDT from vince-q @ Cascade Lodge BBS

[Reply] [ReplyQuoted] [Headers] [Print]

Aug 18 2013 9:04pm from IGnatius T Foobar @uncnsrd (Uncensored)
Ok, that makes more sense -- you are suspecting the VirtualBox software

specifically, not virtualization in general. Uncensored has been
running virtualized for about six years now, originally on Virtuozzo
and then briefly on VMware but mostly (and currently) on KVM.

I actually suspect that the "performance optimizations" made to that

layer will probably have something to do with it, but I will leave that

to dothebart to figure out.



Again, I do not suspect Citadel code to be the culprit here. I simply do not, and never have, trusted "VirtualBox." I have seen such flakiness from that unmitigated POS that I firmly believe it should be banned from all serious networks. If it cannot reliably keep track of the system's **mouse** how can it be expected to keep track of a TCP/IP stack??? Or the hardware bound thereto???

[#] Mon Aug 19 2013 10:25:07 EDT from Freakdog @ Dog Pound BBS II

[Reply] [ReplyQuoted] [Headers] [Print]

 

Mon Aug 19 2013 02:13:14 AM EDT from vince-q @ Cascade Lodge BBS

Again, I do not suspect Citadel code to be the culprit here. I simply do not, and never have, trusted "VirtualBox." I have seen such flakiness from that unmitigated POS that I firmly believe it should be banned from all serious networks. If it cannot reliably keep track of the system's **mouse** how can it be expected to keep track of a TCP/IP stack??? Or the hardware bound thereto???

With all due respect...the VM, itself, never goes off of the network, and on none the virtualized servers or desktops I've run inside of VB, have I ever lost track of the mouse.

In the case of the Citadel VM, I do not make use of NAT. Rather, I have put the virtual interface right onto the network, with its own IP, etc.

There have been no network hiccups since I upgraded both DP2 and Cascade. The only issue I've had was a recurrence of the database related Citadel halt.

In addition, I've stopped polling Cascade directly from DP2...all network based connections are now initiated solely by Cascade.

Additionally, the networking issues that we experienced prior to the upgrade to 8.20 had nothing to do with polling (i.e. TCP/IP based connections between the systems) and everything to do with spooling (import and/or export of the network packets which are transmitted between systems via the polling process). Those issues were not just happening on DP2, but also on Uncensored, which usually required a Citadel restart at either or both ends, and those appear to have been remedied.

I've been running Citadel inside a VM for over a year and a half, now, and so far, all of the issues I've had with Citadel have been identifiable and, so far, patchable by the CitDev team.



Go to page: First ... 63 64 65 66 [67] 68 69 70 71 ... Last