Oversized NAS

[Joeyfla95]

1 minute ago, Joeyfla95 said:

i just want to offload stress on the RAM therefore i am using an LSI card

if i have to i will get a windows server licence

[Xhypno402]

Just now, Joeyfla95 said:

better safe then sorry m9.

better safe then sorry would matter if he had an array with 24-48 TB and had enough data usage to warrant it. spending 50%+ more on ram is a poor usage of money if you won't see an actual benefit from it.

[79wjd]

 

4 minutes ago, Joeyfla95 said:

i just want to offload stress on the RAM therefore i am using an LSI card

There's no reason to take the stress off the RAM, and if you really want to use a RAID card, then you shouldn't use FreeNAS. 

[F4S4K4N]

2 hours ago, Joeyfla95 said:

the raid cards meant to use R5 so IF a drive fails -unlikely- i can replace it easily. and not suffer data loss

For all intensive purposes, ZFS is a RAID card, a RAID card that is smarter than virtually any other hardware RAID card. When you run a hardware RAID card with ZFS you're essentially lying to ZFS by telling it you have one virtual hard drive.

 

Let me tell you something from personal experience, ZFS is like a woman. If you lie to it, and it never finds out, well i guess you're probably okay. However if you lie to it, and it does find out, you're screwed. 

 

2 hours ago, Xhypno402 said:

Actually you are very wrong in all of your statements.

 

1. ECC ram does not increase server stability. It increases data consistence for applications. ECC by no means now or ever has stopped a server from crashing Windows, Linux, BSD, or any.

 

2. Raid cards use ECC because ECC supports low voltage battery discharge while non-ECC does not. This allows highend raid cards to have a battery backup connected to the raid card to allow storage of data in memory till system power is restored to stop the loss of in flight writes.

 

3. The argument for ecc vs non-ecc has been a heated one with the freenas community. They have pushed for a long time that the use of non-ecc memory can kill a pool. This is also a misnomer.

 

ZFS was designed by SUN initially for Solaris on lower end sparc servers that did not use ECC as they wanted you to use their Sun Disk Arrays for primary data storage on their high end servers with ECC ram for application integrity. I worked for many years with large scale deployments of these at Affinity Internet as the backing for their high performance cluster shared hosting environments. The key to why the ECC memory was used with the app servers is with long standing data in memory there is a chance that data decay of a flipped bit can occur. This can be a major issue if the data that is being read is an in memory data base containing things like transnational data records or a static routing table for 100k websites.

 

ECC does not correct for the common cause of data corruption in memory which is multi-bit or word data errors. This is the type of error that will bork a pool or a strip, not a single bit. Yes with read/write throughput from ZFS the chance of memory corruption is higher, but as has been discovered in many test, the amount of ram usage has to be to multiple orders of magnitude higher then a common nas will see. 

 

 

1. If a critical Windows service had some data in RAM, that data experiences a single bit flip, and that server reads that corrupted data, there's a possibility that service could crash and / or blue screen the box. So ECC RAM can increase stability.

 

2. If they wanted to use non-ECC RAM, they could just as easily have it made to support low voltage battery discharge. Take into account RAID5 doing parity check-summing for instance. That checksum immediately goes into RAM. If you have a RAM stick with a stuck bit where that checksum is stored, then you're checksum is corrupted and the RAID card will probably eject the disk.

 

3. It's not going to kill a pool until it kicks the bucket or picks up a stray bit of cosmic radiation in the wrong spot. In non mission-critical applications i run ZFS without ECC and have only had a pool corrupt one time which was when a RAM stick failed. How it failed is up for speculation though. If it just dumped all of it's contents and died, ECC wouldn't save you either.

 

IMO, this is a spot on description of what happens with single bit errors. Whether or not you trust non-ECC ram or not i suppose is a personal choice.

 

https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/

 

2 hours ago, Joeyfla95 said:

i just want to offload stress on the RAM therefore i am using an LSI card

 

Well then you don't want to use FreeNAS or ZFS as the LSI card wont make ZFS use any less RAM than before.

[Joeyfla95]

7 hours ago, F4S4K4N said:

For all intensive purposes, ZFS is a RAID card, a RAID card that is smarter than virtually any other hardware RAID card. When you run a hardware RAID card with ZFS you're essentially lying to ZFS by telling it you have one virtual hard drive.

 

Let me tell you something from personal experience, ZFS is like a woman. If you lie to it, and it never finds out, well i guess you're probably okay. However if you lie to it, and it does find out, you're screwed. 

 

 

 

1. If a critical Windows service had some data in RAM, that data experiences a single bit flip, and that server reads that corrupted data, there's a possibility that service could crash and / or blue screen the box. So ECC RAM can increase stability.

 

2. If they wanted to use non-ECC RAM, they could just as easily have it made to support low voltage battery discharge. Take into account RAID5 doing parity check-summing for instance. That checksum immediately goes into RAM. If you have a RAM stick with a stuck bit where that checksum is stored, then you're checksum is corrupted and the RAID card will probably eject the disk.

 

3. It's not going to kill a pool until it kicks the bucket or picks up a stray bit of cosmic radiation in the wrong spot. In non mission-critical applications i run ZFS without ECC and have only had a pool corrupt one time which was when a RAM stick failed. How it failed is up for speculation though. If it just dumped all of it's contents and died, ECC wouldn't save you either.

 

IMO, this is a spot on description of what happens with single bit errors. Whether or not you trust non-ECC ram or not i suppose is a personal choice.

 

https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/

 

 

Well then you don't want to use FreeNAS or ZFS as the LSI card wont make ZFS use any less RAM than before.

7 hours ago, F4S4K4N said:

so my only option

7 hours ago, F4S4K4N said:

For all intensive purposes, ZFS is a RAID card, a RAID card that is smarter than virtually any other hardware RAID card. When you run a hardware RAID card with ZFS you're essentially lying to ZFS by telling it you have one virtual hard drive.

 

Let me tell you something from personal experience, ZFS is like a woman. If you lie to it, and it never finds out, well i guess you're probably okay. However if you lie to it, and it does find out, you're screwed. 

 

 

 

1. If a critical Windows service had some data in RAM, that data experiences a single bit flip, and that server reads that corrupted data, there's a possibility that service could crash and / or blue screen the box. So ECC RAM can increase stability.

 

2. If they wanted to use non-ECC RAM, they could just as easily have it made to support low voltage battery discharge. Take into account RAID5 doing parity check-summing for instance. That checksum immediately goes into RAM. If you have a RAM stick with a stuck bit where that checksum is stored, then you're checksum is corrupted and the RAID card will probably eject the disk.

 

3. It's not going to kill a pool until it kicks the bucket or picks up a stray bit of cosmic radiation in the wrong spot. In non mission-critical applications i run ZFS without ECC and have only had a pool corrupt one time which was when a RAM stick failed. How it failed is up for speculation though. If it just dumped all of it's contents and died, ECC wouldn't save you either.

 

IMO, this is a spot on description of what happens with single bit errors. Whether or not you trust non-ECC ram or not i suppose is a personal choice.

 

https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/

 

 

Well then you don't want to use FreeNAS or ZFS as the LSI card wont make ZFS use any less RAM than before.

 

7 hours ago, F4S4K4N said:

For all intensive purposes, ZFS is a RAID card, a RAID card that is smarter than virtually any other hardware RAID card. When you run a hardware RAID card with ZFS you're essentially lying to ZFS by telling it you have one virtual hard drive.

 

Let me tell you something from personal experience, ZFS is like a woman. If you lie to it, and it never finds out, well i guess you're probably okay. However if you lie to it, and it does find out, you're screwed. 

 

 

 

1. If a critical Windows service had some data in RAM, that data experiences a single bit flip, and that server reads that corrupted data, there's a possibility that service could crash and / or blue screen the box. So ECC RAM can increase stability.

 

2. If they wanted to use non-ECC RAM, they could just as easily have it made to support low voltage battery discharge. Take into account RAID5 doing parity check-summing for instance. That checksum immediately goes into RAM. If you have a RAM stick with a stuck bit where that checksum is stored, then you're checksum is corrupted and the RAID card will probably eject the disk.

 

3. It's not going to kill a pool until it kicks the bucket or picks up a stray bit of cosmic radiation in the wrong spot. In non mission-critical applications i run ZFS without ECC and have only had a pool corrupt one time which was when a RAM stick failed. How it failed is up for speculation though. If it just dumped all of it's contents and died, ECC wouldn't save you either.

 

IMO, this is a spot on description of what happens with single bit errors. Whether or not you trust non-ECC ram or not i suppose is a personal choice.

 

https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/

 

 

Well then you don't want to use FreeNAS or ZFS as the LSI card wont make ZFS use any less RAM than before.

For all intensive purposes, ZFS is a RAID card, a RAID card that is smarter than virtually any other hardware RAID card. When you run a hardware RAID card with ZFS you're essentially lying to ZFS by telling it you have one virtual hard drive.

 

Let me tell you something from personal experience, ZFS is like a woman. If you lie to it, and it never finds out, well i guess you're probably okay. However if you lie to it, and it does find out, you're screwed. 

 

 

 

1. If a critical Windows service had some data in RAM, that data experiences a single bit flip, and that server reads that corrupted data, there's a possibility that service could crash and / or blue screen the box. So ECC RAM can increase stability.

 

2. If they wanted to use non-ECC RAM, they could just as easily have it made to support low voltage battery discharge. Take into account RAID5 doing parity check-summing for instance. That checksum immediately goes into RAM. If you have a RAM stick with a stuck bit where that checksum is stored, then you're checksum is corrupted and the RAID card will probably eject the disk.

 

3. It's not going to kill a pool until it kicks the bucket or picks up a stray bit of cosmic radiation in the wrong spot. In non mission-critical applications i run ZFS without ECC and have only had a pool corrupt one time which was when a RAM stick failed. How it failed is up for speculation though. If it just dumped all of it's contents and died, ECC wouldn't save you either.

 

IMO, this is a spot on description of what happens with single bit errors. Whether or not you trust non-ECC ram or not i suppose is a personal choice.

 

https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/

 

 

well Sh1t i just wasted 300$ on a paperweight....

 

 

[Joeyfla95]

i hate the quote system

[beavo451]

19 hours ago, Joeyfla95 said:

i hate the quote system

Bottom line.

 

ECC is NOT required for FreeNAS. Just recommended.

 

You cannot make any system 100% fail-proof. You NEED (yes actually need and not just recommended) backups.

[Joeyfla95]

6 hours ago, beavo451 said:

Bottom line.

 

ECC is NOT required for FreeNAS. Just recommended.

 

You cannot make any system 100% fail-proof. You NEED (yes actually need and not just recommended) backups.

my plan is this. i want the nas as fail proof as possible. therefore i use ecc and R5. i have a backup machine consisting of a bunch of 4tb wd blacks. i shall write the red drives to the blacks every week. the old backups shall be overwritten.

[beavo451]

7 hours ago, Joeyfla95 said:

my plan is this. i want the nas as fail proof as possible. therefore i use ecc and R5. i have a backup machine consisting of a bunch of 4tb wd blacks. i shall write the red drives to the blacks every week. the old backups shall be overwritten.

If you use FreeNAS, you cannot use RAID.

 

Do not overwrite backups! You need several copies that go back a few weeks or months in case a file gets changed accidentally or there is a failure during a backup.

 

And if it really is mission critical data, you need to do backups way more often.

[Joeyfla95]

On 3/12/2016 at 7:34 AM, beavo451 said:

If you use FreeNAS, you cannot use RAID.

well then im using windows server. Final..... along with the LSI card

 

[beavo451]

15 minutes ago, Joeyfla95 said:

well then im using windows server. Final..... along with the LSI card

 

What is your fixation on RAID?