This is topic GDC server won't boot with USB drive inserted in forum Digital Cinema Forum at Film-Tech Forum ARCHIVE.


To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=16;t=001226

---

Posted by Steven J Hart (Member # 2340) on 09-24-2012, 10:08 AM:
 
Just a heads up for those of you with GDC servers:
We had an extended power outage at our location early Sunday morning. The battery on our UPS backup was depleted so the server lost power. When my staff went to restart the server it wouldn't boot up. When I got to the theater, I noted that I'd forgotten a USB drive in the server. It booted right up after pulling out the USB drive.

Steve
 

---

Posted by Monte L Fullmer (Member # 2797) on 09-24-2012, 12:16 PM:
 
Bet one can go into the BIOS and change the boot priority to eliminate this.
 

---

Posted by Dennis Benjamin (Member # 1137) on 09-25-2012, 09:58 AM:
 
Not unless you have the BIOS password.....
 

---

Posted by Chad Souder (Member # 343) on 09-25-2012, 07:36 PM:
 
This is the common boot sequence for most PCs. Leaving an ipod plugged in can cause the same problem. It's done that way for a reason - something to do with software upgrades I think. A computer guy could tell you better why.
 

---

Posted by Brian Shull (Member # 4585) on 09-26-2012, 12:23 AM:
 
It was a built in safety for years on computers. If your operating system fails you could in theory load a operating system on the USB and boot the computer to try to fix the problem without loosing data. As said you could change the boot sequence in the BIOS if you can access it, and password.

Brian
 

---

Posted by Ken Lackner (Member # 1002) on 09-26-2012, 09:01 AM:
 
What theory? That's exactly why PC's are designed this way. Back in the days of DOS and floppy disk drives, if the operating system on your C: drive failed (or if the hard drive itself failed) and you could not boot your PC, you'd insert a boot disk in the A: drive and reboot the system. The boot priority was such that it would first look to the A: drive for a bootable OS. If no disk was loaded in the A: drive, it would then look to the C: drive. If a disk that was not bootable was in the A: drive, you would receive an error message along the lines of "Non system disk or disk error. Replace and strike any key when ready." This tells the user that there is no bootable OS loaded on the drive, therefore the system cannot boot. You would receive the same error message if there was no disk in the A: drive but your C: drive, or the OS on it, was bad.

Booting from the A: drive was also how a new OS was installed to a blank (therefore non-bootable) HDD. A disk containing just enough files to booth the system and get to a command prompt would be loaded in the A: drive; the system would be booted up, and the user would then install the new OS from the A: drive, or even the CD-ROM drive if CD-ROM drivers were loaded onto the boot disk.

The BIOS on modern systems looks at the USB ports for a bootable system drive the same way older machines looked at the A: drive before the C: drive.

Having no first-hand experience with GDC, I cannot tell you whether they have the boot priority set that way for a specific reason; you most certainly should check with them before changing it.
 

---

Posted by Brad Miller (Member # 2) on 09-26-2012, 01:12 PM:
 
There is no reason to have the boot priority look to the USB. With digital cinema, there are too many instances of when people NEED to ingest via USB, and that means people WILL leave a USB drive plugged in from time to time. The designers should've thought about this. Having the boot priority look to the CD makes sense for such necessary situations should they arise, but not USB.

What's more foolish is not having a server wake-up upon seeing AC power.

Servers can be configured most any way possible, but in some instances you are at the mercy of the designer understanding the market that the product is being used in.
 

---

Posted by Tony Bandiera Jr (Member # 2365) on 09-26-2012, 03:38 PM:
 
Seems that GDC has missed the mark in many ways..and I heard a rumor that it is (was?) possible to revive an expired key by changing the BIOS date and time...oops.

I get the impression that the designers at GDC have never been in, much less operated, a cinema.
 

---

Posted by Gordon McLeod (Member # 33) on 09-26-2012, 04:36 PM:
 
From there web site

"Dr. Chong Man Nang is the founder and the CEO of GDC Technology.
Some of his accolades includes:




Leading the team that won the first prize of the worldwide Texas Instruments' Digital Signal Processing Solutions Challenge US$100,000 award (1996)


Spearheading the architecture, design, and successful commercial release of Nirvana’s award-winning REVIVAL DIGITAL® product lines used by Hollywood studios for the digital restoration of classic feature films.


Serving as the chief consultant of Da Vinci Technologies (2000-2002)


Establishing GDC Technology in 1999 and developing it into a world leading digital cinema solution provider."

Also I know that currently you need a key to alter the time on the server
 

---

Posted by Tony Bandiera Jr (Member # 2365) on 09-26-2012, 04:55 PM:
 
Gordon, those appear to all be pre- and post-production credentials so I stand by my original statement that they haven't a clue about real-world, exhibition operations.
 

---

Posted by Gordon McLeod (Member # 33) on 09-26-2012, 05:01 PM:
 
All I know we have had no complaints yet with GDC servers we have been installing
 

---

Posted by Adam Martin (Member # 641) on 09-26-2012, 06:29 PM:
 
Is your phone working?
 

---

Posted by Brad Miller (Member # 2) on 09-26-2012, 09:34 PM:
 
Marketing babble Gordon.

So Gordon, what version of the software are you running?
 

---

Posted by Steven J Hart (Member # 2340) on 09-26-2012, 10:38 PM:
 
I had "No Major Complaints" about the GDC until last weekend when we MISSED A SHOW because the server would not reboot after a power outage
 

---

Posted by Brad Miller (Member # 2) on 09-26-2012, 11:27 PM:
 
Well Steven, maybe. What version of software are you running? GDC has been touting that their servers are DCI-approved since the first of the year, but come to find out they are just now releasing the actual DCI-approved software. (Highly misleading marketing and sales efforts there, because without this software the servers are NOT DCI-approved.)

The deadline for upgrading to this software is October 31st or VPFs are forfeited.

Here's the kicker...when you power cycle the server, expect a 20 MINUTE reboot time now! I've timed this and it was 22 minutes before I could play content again. It's an absolute showstopper. Time to stock up on emergency passes.

Could it be you are upgraded to the DCI compliant software and that was all it was, and the server worked fine the next show?

The amusing part of this is there were a few people that would always complain that Dolby servers took a few extra minutes to reboot over GDC and Doremi. Now that everyone is on the same level playing field with DCI-compliance, Dolby servers boot the fastest...still at their typical 7 minute reboot time. Doremi is somewhere around 10 minutes and GDC drags it up the end at 20 minutes.
 

---

Posted by Tony Bandiera Jr (Member # 2365) on 09-26-2012, 11:32 PM:
 
This thread points out some of the main things brought up in the "Best Server" thread....but can be summed up in one well-worn phrase:

You get what you pay for.

I wonder how many theatres with GDC's have had the "savings" wiped out in lost shows and/or refunds?
 

---

Posted by Brad Miller (Member # 2) on 09-26-2012, 11:38 PM:
 
But that's just it Tony. GDC servers are NOT the cheapest. Sure at one time they were, but for the last year Dolby and Doremi have been kicking GDC in the face with their price.

My personal opinion...it was due to the GDC VPF program. They knew they could put any price they wanted to on their servers and people HAD to buy them! And I'll back that up by saying let's see what happens in the next 6 months once they have moved all of the servers they can. I'll bet $20 that their price will magically drop down drastically again.
 

---

Posted by Tony Bandiera Jr (Member # 2365) on 09-26-2012, 11:53 PM:
 
Thank you for the clarification Brad. I have noticed and been enjoying the price drops. Maybe someday soon I can afford a D-cinema setup for my palatial estate. Then my journey to the Dark Side will be complete.
 

---

Posted by Joe Elliott (Member # 4006) on 09-27-2012, 01:30 AM:
 
I just updated a few to 8.01, and they took just under 5 minutes to boot up. I have heard it will take over 20 minutes if you shut them down wrong (via the power button instead of the menu). You might want to check back on those servers Brad, sounds like there is something wrong.
 

---

Posted by Brad Miller (Member # 2) on 09-27-2012, 01:45 AM:
 
We were talking about power failures Joe. Yes in theory they don't boot up that slow every time, but if the thing is acting screwy a software reboot won't always work and the power cycle becomes the only option. Regardless, a power failure on a GDC will of course guarantee you that 20 minute intermission. Why is this considered acceptable? It certainly isn't by my standards.

To put it in Joe-like terms, just imagine the fun at a 16 plex with a power failure. Seriously, put yourself in anyone's shoes:
Manager's position...sucks to be you.
Projectionist's position...sucks to be you.
Patron's position...sucks you patronized that theater.

The DCI people need to step in and make a MANDATORY requirement that all servers, projectors and anything else in the chain preventing the show from running must be able to boot from ANY condition in X number of minutes. (My vote is a maximum of 5, but 3 is preferred.)
 

---

Posted by Joe Elliott (Member # 4006) on 09-27-2012, 02:14 AM:
 
Sorry, I was going off of your own statements that sounded as if it was booting that way every time.

quote: Brad Miller

---

Dolby servers boot the fastest...still at their typical 7 minute reboot time. Doremi is somewhere around 10 minutes and GDC drags it up the end at 20 minutes.

---

quote:

---

Why power outages are bad for your data
1. Introduction

A lot of people assume that when power is cut from their computer, it should be able to handle that gracefully and no data should be lost. Unfortunately, it is not that simple. For a lot of people the risk is low, but when you're using your computer even slightly more seriously and your data is equally important, you might want to consider using a UPS.

This article explains what happens to the hardware and software when the power fails and the possible consequences it has. It should give you enough insight to decide whether you need a UPS or not.
2. What happens hardware-wise
2.1. Direct result

When the power fails, no individual component gets a clean shutdown command; power is just removed. When this happens, some parts of the machine may last longer than other parts. One of the first things that will happen, is that the memory DIMMs will no longer be refreshed properly (DRAM needs to be refreshed constantly otherwise it will lose its data) and very rapidly, the memory will contain only garbage. The hard drives and DMA controller however, will run a bit longer; so if data is being written to disk, the DMA controller will keep reading data from memory, but it has no idea that this data is corrupted. Some file systems are more sensitive to this kind of failure, because of the different kinds of journaling they do.

There are certain machines which are protected against this type of data corruption, by having the power supply send an interrupt to the operating system when power fails, but ordinary class PC hardware does not.

There is another side to this story, however. Researchers have shown that encryption keys can be retreived from memory minutes after the computer has been shut down. This would suggest that the memory doesn't corrupt so quickly at all when power is removed. However, because machines are still unstable when you decrease the refresh cycle of your RAM, it is apparent that some corruption still occurs.
2.2. Indirect result

Not directly related to a power loss as much as any kind of shutdown, is the fact that hard drives which have reached a certain age tend to die when the machine is powered up again. This doesn't necessarily happen immediately, but can take a few days. Additionally, you can imagine what will happen in this scenario when you have an array of identical disks of equal age; always put spare disks in your array, in power safe mode so they don't wear down.
3. What happens software-wise
3.1. Disk cache

Disk write cache is used to collect and delay transfers to the disks in favor of speed, because memory is faster than disks. When you shut down a machine when there is uncommitted data in the cache, you will lose this data, or corrupt it because only part of the cache is written. This can be illustrated very nicely by booting your Linux machine with the kernel parameter "init=/bin/bash". This will start a shell instead of the initialization procedure. You can then edit files, like /etc/shadow, should you want to reset your password. If you then press ctrl-alt-del without running the "sync" command to commit the disk cache first, your changes will not be committed to disk.

There are different kind of cache systems in existence. Two important ones for write cache are write-through and write-back. The former is safe, because it reports the data as written when the data is committed to disk. The latter is unsafe, because it reports the data as written when it has been written to cache, while it hasn't been written to disk yet. Even when you have a UPS it's unsafe, because there are several other reaons when power can suddenly disappear.

Write-back cache is used in a lot of disks these days. If possible, you may want to consider turning it off.
3.2. (Encrypted) file systems

Most people will think that because of journaling, file systems are protected against power failures. It's true that filesystems with journaling are more robust than those without, but it should be clear by now there are some things the file system cannot protect against.

Then of course, there are different ways that journaling can be done. Ext3 is more resillient against power failures then XFS and ReiserFS, because ext3 does physical block journaling. Ext4, however, by default does something called "delayed allocation," which means that meta-data is saved more often than the data itself. In the event of a power failure, this can data corruption, for obvious reasons. Linux Torvalds has an outspoken opinion against delayed allocation.

As the Gentoo Wiki states, you are even more susceptible to data loss in the event of a power failure when using an encrypted file system. The reason for this, is that hard disks are block devices. Normally, if a few bits are flipped as you write it to disk, you can take advantage of properties inherent in the data (depending on what it is) to recover it. At the very least, you only lose a bit of the block. Should the block be encrypted and trashed as it is written, the flipped bit will cause decryption of that block to fail, so it is all lost. Entropy of most computer data on disk is quite low, while encrypted data is essentially indistinguishable from true random; guessing gets a lot harder. Additionally, it is possible for multiple subsequent blocks to also be lost due to the initialization vectors used in IV chaining being unrecoverable. This depends on whether or not you're using them, how the cipher is configured, and other factors, but is a consideration.
3.3. (Linux software) RAID

Linux software RAID, and any RAID basically, needs to know if the disks of the array are still properly matched to eachother when the array is initialized. When power fails, or when you press reset, they will be in a "dirty" state, and the system may need to recreate the array. That is, if it can. I've never tried it, but I can imagine that a RAID0 can be completely destroyed by a power failure. But, don't take my word for that...

Modern Linux kernels (2.6.16 and newer) and raid tools (mdadm 2.4.1 and newer) luckily have a precaution against that, namely a write intent bitmap. When using Linux software RAID, I'd advise you to enable this. There are enough resources on the internet where you can find how, like the Gentoo Wiki. Unfortunately, write intent bitmaps are very slow. When you have the protection of a UPS, not using a write intent bitmap becomes more acceptable.
3.4. Databases

When you use a database system with (good) transaction support, data corruption will not happen when the power is removed. That is, when you're using disks that don't lie about their cache status (see above). You will, of course, lose all the uncommitted transactions, which can be annoying in itself.

When you're running a database without transaction support, like MySQL MyISAM, data corruption is of course likely.

Applications like LDAP directories, source control management repositories, etc, are also potentially susceptible to the same kind of failures, depending on if and how well they use transactions. Like I describe in my backup article, it's also important to make scheduled dumps of such applications, to make sure have a backup in a robust, self-contained archive.
4. Surge protection

UPSes also protect against surges on the mains power. However, only an online UPS (the expensive sort) does it properly, by always running the load from the battery. The offline variety merely uses MOVs, which is exactly the same thing as those ordinary power strips with surge protection. The effectiveness of those things can be questioned because of delay time, impedance (resistance) of the safety earth, longetivy of the MOVs because of frequent surges, etc. The internet is filled with information about the fallibility of MOV surge protection.

The surge protection on UPSes also often includes protection for ethernet and/or telephone networks. I really advice against using those. When there is a surge, the MOVs temporarily short the line containing the surge with the safety earth, but it will also connect the data networks to it. This safety earth, however, does not have infinitely low impedance, and therefore it's possible that some of the excess current will travel up the network, as opposed to down the safety earth. The exact details of this are more complex than this, but as always, the internet is your tool should you want to find out more.

---

quote: Brad Miller

---

Could it be you are upgraded to the DCI compliant software and that was all it was, and the server worked fine the next show?

---