| To: | Rick Spillane <necro351@xxxxxxxxx> |
|---|---|
| Subject: | Re: ACID Question |
| From: | Andi Kleen <ak@xxxxxx> |
| Date: | Mon, 07 Feb 2005 20:04:50 +0100 |
| Cc: | linux-xfs@xxxxxxxxxxx |
| In-reply-to: | <ccdab71705020709446c578e99@xxxxxxxxxxxxxx> (Rick Spillane's message of "Mon, 7 Feb 2005 12:44:13 -0500") |
| References: | <ccdab71705020709446c578e99@xxxxxxxxxxxxxx> |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
| User-agent: | Gnus/5.110002 (No Gnus v0.2) Emacs/21.3 (gnu/linux) |
Rick Spillane <necro351@xxxxxxxxx> writes: > Does XFS entirely follow the ACID requirements: Atomic, Consistancy, > Isolated, and Durability? For user data (read/write) etc. no. It follows POSIX semantics for that, which are much weaker. You can implemented full ACID on top of it though using fsync/fdatasync/O_SYNC. It probably does for internal transactions on its disk data structures (like updating directories or inodes), but user visible actions (rename, create file etc.) are also not necessary atomic or isolated because the Linux VFS often splits them up into multiple actions. They should be all consitent in itself though and durable if you flush the data at some point (fsync). -Andi |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Delayed Allocation Question, Sonny Rao |
|---|---|
| Next by Date: | Re: Delayed Allocation Question, Steve Lord |
| Previous by Thread: | ACID Question, Rick Spillane |
| Next by Thread: | Re: ACID Question, Stewart Smith |
| Indexes: | [Date] [Thread] [Top] [All Lists] |