I think I’ve got a semi permament installation coming up for a building in Birmingham, got 1 more meeting before I know for sure, but 1 question I have still to sort out in my own head is how to back it up?
Its a fairly simple install a media server playing a variety of files some random and some scheduled, I need to guarentee as close to 100% up time that I can, I’m going to include a ups in case of power loss, and leave the system permantly running, in tests I’ve had no crashes so far, but obviously I cant guarentee that!
Im thinking of either having a second pc or a dvd running as well and if there a problem a switch they can hit to change over, although an automatic way would be better! Its local so I can go down and sort it out, but I work away from home quite regulary too!
So how do meso do it?
Thanks again
make sure that your pc and other hardware resumes to the same state as they where before, if you switch the power off. there is no more you can do, that means no more reasonable things in my opinion… put your root patch to the autostart with the fullscreen pin of the renderer to 1, disable all windows notifications (with xp-antispy or so), set the bios options to auto on after power fail, and any other auto power on functions. your customer should be able to switch everything on and off with the power plug…
Mmmm, on previous times I’ve left patches unattended, and had someone switch it on (and off more importantly!) I’ve had phone calls saying the outputs just flashing (an error I could never reproduce, despite generators being turned off before shutdown when I was there), or other strange errors corrected by overwriting the patch with a backup version. Its partly the reason I’ve always wanted a way of solidifing a patch and locking it into an exe (of course that can corrupt too but its 1 file that has to be replaced!)
Other problems with bumping the power has been disk errors, but this is with video files playing.
I’ve never come across as many errors as when I leave a machine somewhere, Even windows has gone tits up on occasion for no reason!
So generally, you just rely on 1 machine and no backups?
Trial period for this install is approx 6 months, with a possible permament install after this, I just want as few call outs as possible, I find the phone calls very stressfull!
for the mb trade fair exhibits there is one spare pc in case one crashes and every one has all files of each other. during the installation we switched the power often on and off (also with video playing) and i’v never seen a single error (the screens had more problems). on tour there will be no other on/off option. just make sure that the pc is clean, windows, vvvv and needed tools, not more. … and of course that you have a backup at home…
This means, that you save the state of a perfectly running system and resume this on every reboot. instead of going throgh the normal boot process of windows (which has the risk of strange window-focus problems of the fullscreen renderer), a memory image is written to disk, then reloaded and resumed as if the system was never down.
There is just one problem: vvvv doesn’t resume rendering after resuming from hibernation. This is a known bug. I don’t know if this bug has been fixed but if not, it should be high priority…
just had an idea about an automatic fall-back solution. let the main machine send the framerate, maybe some color values aquired with pipet to the fall-back machine via UDP. if the data is wrong or not being send anymore the fall-back machine jumps in, does the switch itself and keeps on displaying the content. and finally restarts the main-machine until the UDP data is fine again. it could actually send a mail to u, to keep u informed about the failure.
anyway, i never seen vvvv crash once it’s working. hope u dont encounter to much trouble with your installation.
@u7
Thats kind of what I was thinking about, an rs232 matrix switching the video if the one fails, a send mail option would be good there too, how would I go about that? Have to say I hadnt thought about an automatic restart of the aflicted machine, that could be clever!
Im hoping for no crashes, I just know the client wants backups! I have had occasional crashes with vvvv, but only when running heavy patches with large spreads, which this one isnt, so I’ve got my fingers crossed!
By the way caught the opening of the george michael tour on sky news this morning, nice wide shot from the back, of his entrance, so to speak, looked very nice! You got any more footage from those shows?
u could send a mail via a php form using html renderer. i’m not a php expert but i’ve seen many tutorials online to set up something like this pretty easy. on the other hand having a network connexion and internet could cause many problems…so best is having a firewall blocking everything apart from port 80 outwards and the udp port inwards.
the automatic restart could be done with a relay and arduino board by switching the reset button. i’d really like to built a fallback system…just my machines never crash ;)
i agree that a backup machine is an absolute must for big productions (best is having a third machine to be able to swap parts with). but this is almost only due to hardware issues - as soon as vvvv is actually running, we are usually very confident it will continue to do so. we have trade fair exhibits and media installations running for weeks and months without problems. Whenever possible i´d schedule a periodic reboot (like every day or every week).
And let me add that at the george michael show the DVI switching matrix for switching between the main and the backup machine introduced a whole bunch of quirks which made the whole thing much more instable than just one machine running vvvv. you are more than doubling the complexity here. although you can handle phone calls somehow more relaxed, you will get more than twice as many phone calls - and every professional who is worth his money will get you to repair the backup IMMEDIATELY anyway.
i guess an automatic handover would open up another can of worms - i doubt it would really help more than it might hurt.
anyway we usually send out periodic UDP broadcasts (containing system status) from every machine, so you can detect if some of them are missing.
Just running a test patch to test stability, I got a
the instruction at “0x7c910543” referenced memory at “0x00000009” The memory could not be read
click ok to terminate.
I dont know whether this is memory defragmentation or a memory leak, but after running for a couple of days changing avi’s every 2 seconds, the memory node reports that the amount of available memory is dropping, its no where near as fast as qtime alternative used to.
I dont know if this is expected behaviour or bug? Anyway I think I can get away with an automated weekly reboot, hopefully everyting will hold together! Depending on the wind conditions the screen should be being installed today for lighting up tomorrow, I’ll post some pictures when it is!
“I dont know if this is expected behaviour or bug?”
hard to tell. we’ve never had an installation doing anything similar. doesn’t it stutt, when you change avis?
btw. have you tried the latest qt-alternative? does it still show this problem?
anyway good luck for your thing. looking forward to the pictures.
In Ricos we have every machine running a RAID in mirror, so if a harddrive crashes… vvvv never failed me in Rico’s once it was running!! (only a shit load off COMport device conflicts, but that is just windows). Someone (thought it was Joreg) made a checklist once, on how to set your machine… stuf like: set your Bios at always boot, even if mouse/keyboard is missing, but can’t find it in the Wiki atm :(
