1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
|
(checksum...) failed
fsck foo (fixing location log)
Only 1 of 2 trustworthy copies exist of foo
Back it up with git-annex copy.
> You've given me severely partial output, and no test case, but until
> it says "fsck foo", the output is pertaining to some other file than foo.
> As far as I can see, there is no bug here. --[[Joey]]
>> Sorry, I thought it would be obvious, but that's no excuse for not
>> providing additional explanation. The problem is that fsck tells me a
>> file's fsck has failed without printing extra details. In this case, the
>> checksum is OK while I don't have enough copies to satisfy the fsck. The
>> fact that I don't have enough copies is obviously relevant, but I would
>> still like to know if the checksums are OK. -- Richard
>>> I think you're misreading the truncated output you posted. The actual,
>>> full output would make much more sense. --[[Joey]]
>>>> No. I have a total of 14908 annex keys, 3333 of which are on a remote. The only message other than 'checksum OK' and the above is 'git-annex: 11577 failed'.
>>>> I checked several files manually, their checksums are OK so `git annex
>>>> fsck` is reporting those files as completely failed when they "only" miss copies. -- Richard
>>>>> fsck considers not enough copies to be a failure condition; it prints
>>>>> error messages about it etc. That has nothing to do with checksums.
>>>>> --[[Joey]]
>>>>>> I get that. Still, I think it would be _extremely_ useful to know what failures occurred, exactly. Not having enough copies is Not Good, yet not having enough copies and a locally correct file is _lot_ better than having not enough copies and a broken file. I.e. I would prefer:
(checksum...) OK
Not enough copies: Only 1 of 2 trustworthy copies exist of foo
>>>>>> or similar and at the end
git-annex: 0 wrong checksums
git-annex: 11577 with too few copies
>>>>>> In the end, it comes down to the distinction of different failure classes. -- Richard
>>>>>>> For the third, and final time:
>>>>>>> # You are misreading the truncated output you posted
>>>>>>> The "checksum" line is regarding **different** file than the
>>>>>>> not enough copies message. fsck does not attempt to checksum a file
>>>>>>> that is not present. [[done]] --[[Joey]]
>>>>>>>> I realized early on that I pasted the wrong cross-passage, but as there is a ton of the same output, I didn't think it would matter. I wasn't aware that it does not try to checksum when there aren't enough copies. To be fair, you only just mentioned that.
>>>>>>>> Personally, I think that's a bug as it makes ensuring local correctness before copying a file to remotes impossible.
>>>>>>>> Either way, I really didn't know it actually _skipped_ checksumming; that part was missing.
>>>>>>>> For the benefit of anyone else who might read this, this is the correct order:
fsck foo (fixing location log)
Only 1 of 2 trustworthy copies exist of foo
Back it up with git-annex copy.
(checksum...) failed
>>>>>>>> If you would like to keep things this way, fine. I think it's less than ideal, but I don't want to argue, either. -- Richard
|
