Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 24.07.2014 03:38, Robinson, Eric wrote: > On 23.07.2014 23:42, Robinson, Eric wrote: >>> Sounds like you want to run Raid1 (drbd) over Raid0 (ssd+md?). This >>> is more fragile than Raid0 over Raid1, so you might consider >>> implementing Raid0 using DM over multiple drbd devices, each mirroring a single ssd. >> >>> Andreas >> >> >> Our plan is to implement Raid0 using Intel hardware RAID controllers on each server, then Logical Raid1 would be provided by DRBD between the servers. My thought was that if anything happens to a RAID volume, DRBD would simply detach and start using the volume from the other server. No? >> > > Yes, but there are two issues with this: > a) Once you replaced the failed drive you have to re-sync the entire Raid-0. > b) If a single drive fails on system b before you replaced and re-synced system a your data is gone. > > If on the other hand you build multiple DRBD mirrors with a Raid-0 on top: > a) you only have to resync a single drive which is faster and doesn't impact the performance of the other mirrors. > b) you only lose data if the other drive of the degraded mirror fails. > If any other drive fails your data is still fine. > > So while the "Raid-0 over multiple DRBD mirrors" setup is a bit more finicky to handle it provides significantly higher reliability and faster rebuild times. > > > --- > > That sounds like a very interesting solution, but I assume you are talking about using OS software RAID to build the Raid0 over multiple DRBD mirrors. In that case, don't you lose the benefit of the RAID controller's 1GB cache? And, more importantly, aren't you making yourself vulnerable to data corruption caused by power loss (because the cached data is not backed by the controller battery), kernel panic, or other OS issues? If you don't utilize the cache than you cannot lose the cache's contents through a power outage or other failure so that isn't really a problem. However you should be able to just pass the disks through individually if the controller is decent. In that case you'd still benefit from the battery backed cache. Regards, Dennis -- Well... let me see if I understand. My controller (Intel RS25AB080) allows me to create 6 single-drive arrays (technically, each would be a Raid0 'array' with only 1 drive). This would present itself to the OS as 6 separate physical drives, /dev/sda through /dev/sdf. I could then build 6 separate DRBD mirrors, /dev/drbd0 through /dev/drbd5 over the physical devices. Then I would use Linux MD RAID to make one big Raid0 from those 6 drbd devices. Then I would partition the big Raid0 volume. Is this what you are suggesting? --Eric Disclaimer - July 24, 2014 This email and any files transmitted with it are confidential and intended solely for Dennis Jacobfeuerborn,drbd-user at lists.linbit.com. If you are not the named addressee you should not disseminate, distribute, copy or alter this email. Any views or opinions presented in this email are solely those of the author and might not represent those of Physicians' Managed Care or Physician Select Management. Warning: Although Physicians' Managed Care or Physician Select Management has taken reasonable precautions to ensure no viruses are present in this email, the company cannot accept responsibility for any loss or damage arising from the use of this email or attachments. This disclaimer was added by Policy Patrol: http://www.policypatrol.com/