In sync if a silent error is detect marks the block as bad like in scrub.

This commit is contained in:
Andrea Mazzoleni 2013-10-03 22:16:29 +02:00
parent 1f0da860c7
commit 7014486bb7
3 changed files with 26 additions and 11 deletions

View file

@ -10,6 +10,8 @@ SnapRAID HISTORY
up to 89 disks with Hexa (6) parity is possible, but not yet implemented.
Note that more general approaches without these limitations are also
possible, but they would affect badly the performance.
* In sync, if a silent error is found, directly mark the block as bad
just like in scrub, and continue with the sync process.
* Sort files by inode when listing the directory. This improves
performance in scanning.
* Removed the detailed list of errors from the screen output.

6
TODO
View file

@ -5,12 +5,14 @@ This is the list of TODO items for SnapRAID.
- Next
* In sync if a data error is found, we can mark the block as bad, like in scrub
and allow sync to continue.
* Remove/Reduce the opt.gui uses. We can now always output to stdlog becase
it never go to screen.
* Detect files moved from one disk to another. We won't be able to reuse
parity, but we can reuse the hash to check that the file was moved
correctly.
We can assume a file moved to another disk if it has the same name, size,
and timestamp. Directory and inode can be different.
* Add a way to list the content of the array. A new "list" command ?
Something similar at "diff" and "dup".

29
sync.c
View file

@ -122,7 +122,8 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
for(i=blockstart;i<blockmax;++i) {
int one_invalid;
int one_valid;
int skip_this_block;
int error_on_this_block;
int silent_error_on_this_block;
int parity_needs_to_be_updated;
int ret;
snapraid_info info;
@ -171,7 +172,8 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
--autosavemissing;
/* by default process the block, and skip it if something go wrong */
skip_this_block = 0;
error_on_this_block = 0;
silent_error_on_this_block = 0;
/* if we have to use the old hash */
info = info_get(&state->infoarr, i);
@ -245,7 +247,7 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
/* if the file is missing, it means that it was removed during sync */
/* this isn't a serious error, so we skip this block, and continue with others */
skip_this_block = 1;
error_on_this_block = 1;
continue;
} else if (errno == EACCES) {
fprintf(stderr, "No access at file '%s'.\n", handle[j].path);
@ -279,7 +281,7 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
/* if the file is changed, it means that it was modified during sync */
/* this isn't a serious error, so we skip this block, and continue with others */
skip_this_block = 1;
error_on_this_block = 1;
continue;
}
@ -307,11 +309,17 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
if (block_has_updated_hash(block)) {
/* compare the hash */
if (memcmp(hash, block->hash, HASH_SIZE) != 0) {
fprintf(stdlog, "error:%u:%s:%s: Data error at position %u\n", i, handle[j].disk->name, block_file_get(block)->sub, block_file_pos(block));
fprintf(stderr, "Data error in file '%s' at position '%u'\n", handle[j].path, block_file_pos(block));
fprintf(stderr, "DANGER! Unexpected data error in a data disk, it isn't possible to sync.\n");
printf("Stopping to allow recovery. Try with 'snapraid -s %u check'\n", i);
fprintf(stderr, "DANGER! Unexpected data error in a data disk! The block is now marked as bad!\n");
fprintf(stderr, "Try with 'snapraid -e fix' to recover!\n");
/* silent errors are very rare, and are not a signal that a disk */
/* is going to fail. So, we just continue marking the block as bad */
/* just like in scrub */
++silent_error;
goto bail;
silent_error_on_this_block = 1;
continue;
}
} else {
/* if until now the parity doesn't need to be update */
@ -343,8 +351,8 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
}
}
/* if we have read all the data required, proceed with the parity */
if (!skip_this_block) {
/* if we have read all the data required and it's correct, proceed with the parity */
if (!error_on_this_block && !silent_error_on_this_block) {
/* updates the parity only if really needed */
if (parity_needs_to_be_updated) {
@ -402,6 +410,9 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
/* we are also clearing any previous bad and rehash flag */
info_set(&state->infoarr, i, info_make(now, 0, 0));
}
} else if (silent_error_on_this_block) {
/* set the error status keeping the existing time and hash */
info_set(&state->infoarr, i, info_set_bad(info));
}
/* mark the state as needing write */