Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
The DRBD documentation, along with every tutorial I can find, recommends configuring volumes within a resource using sd* names (e.g. /dev/sda). However, I am concerned about the volatility of these names between reboots and during drive replacement. Since I cannot find any references online worrying about device name volatility and DRBD, I am starting to believe that this is not an issue (i.e. DRBD handles this). But, I would love to have confirmation. We are building a pair of storage servers (a primary and a secondary). Each one has a chassis that supports 24 hot-swappable SAS drives. The controller will be configured in IT mode with no RAID. Each drive will be replicated to a corresponding drive on the secondary server. Consider an example: I have two hard drives in each server. The first drive is named sdb and the second is sdc. But, one of the servers reboots. It is possible (albeit unlikely) that the kernel might switch the drive names (i.e. sdb is renamed sdc and vice versa) screwing up the pairing. To eliminate this problem I could configure DRBD with volume IDs, file system UUIDs (although this might not work since DRBD would be using raw devices), by path (this probably will not work with SAS IT), or use udev to name the drives (e.g. by slot). Is this a problem or does DRBD track volumes by ID internally? Does anyone have any recommendations with managing a large quantity of hot-swappable drives (e.g. tracking which drives in which bay, managing the DRBD volume pairings, etc.)? Thank you very much for your time, Ted -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20120123/c21912bb/attachment.htm>