[s3ql] swift (using ovh) backend fails to create usable fs

[Florian Ludwig]

Hi there,

I am trying to use s3ql 2.29 with swift backend. I always end up with a
corrupted backend. (logs below) Downloading the files created by mkfs and
using local:// as a backend also does result in the same error message so I
assume the upload part is the problem, not the downloading part.

Any ideas what could be the culprit?

Thank you very much in advance!
Florian


*# mkfs.s3ql --authfile cred --force --backend-options
disable-expect100,no-feature-detection
swiftks://auth.cloud.ovh.net/DE1:s3ql-test*
Before using S3QL, make sure to read the user's guide, especially
the 'Important Rules to Avoid Losing Data' section.
Purging existing file system data..
Please note that the new file system may appear inconsistent
for a while until the removals have propagated through the backend.
Enter encryption password:
Confirm encryption password:
Generating random encryption key...
Creating metadata tables...
Dumping metadata...
Dumping metadata...
..objects..
..blocks..
..inodes..
..inode_blocks..
..symlink_targets..
..names..
..contents..
..ext_attributes..
Compressing and uploading metadata...
Wrote 154 bytes of compressed metadata.
Cycling metadata backups...
Backing up old metadata...
Please store the following master key in a safe location. It allows
decryption of the S3QL file system in case the storage objects holding
this information get corrupted:
---BEGIN MASTER KEY---
Jkud 6wSU Y4UB BFj0 TMFU ixz0 JI2n 0Dcf RNnY jT/5 w/M=
---END MASTER KEY---

*# mount.s3ql --authfile cred --backend-options disable-expect100
swiftks://auth.cloud.ovh.net/DE1:s3ql-test /test/container/*
Using 8 upload threads.
Autodetected 1048520 file descriptors available for cache entries
Detected Swift features for storage.de1.cloud.ovh.net:443: copy via COPY,
Bulk delete 1000 keys at a time, maximum meta value length is 255 bytes
ERROR: Backend data corrupted, or file system revision needs upgrade.

[Florian Ludwig]

Interestingly it does work without encryption (--plain).

[Nikolaus Rath]

Post by Florian Ludwig
Hi there,
I am trying to use s3ql 2.29 with swift backend. I always end up with a
corrupted backend. (logs below)

My guess is that the backend does not store/return object metadata
correctly.

Post by Florian Ludwig
Downloading the files created by mkfs and
using local:// as a backend also does result in the same error message

The file format is different, this can't work.

Post by Florian Ludwig
so I
assume the upload part is the problem, not the downloading part.

No, it may be either.

Post by Florian Ludwig
Any ideas what could be the culprit?

You could try to uncomment the catching of the exception in
src/s3ql/common.py. This will not fix the problem, but may narrow it
down. Comment out the following lines:

diff --git a/src/s3ql/common.py b/src/s3ql/common.py
--- a/src/s3ql/common.py
+++ b/src/s3ql/common.py
@@ -279,9 +279,9 @@
except DanglingStorageURLError as exc:
raise QuietError(str(exc), exitcode=16)

- except CorruptedObjectError:
- raise QuietError('Backend data corrupted, or file system '
- 'revision needs upgrade.', exitcode=17)
+# except CorruptedObjectError:
+# raise QuietError('Backend data corrupted, or file system '
+# 'revision needs upgrade.', exitcode=17)

except NoSuchObject:
encrypted = False
@@ -309,16 +309,16 @@
compress = getattr(options, 'compress', ('lzma', 2))

with ComprencBackend(fs_passphrase, compress, backend) as tmp_backend:
- try:
+# try:
if encrypted:
data_pw = tmp_backend['s3ql_passphrase']
else:
data_pw = None
# Try to read metadata to detect old file system revision
tmp_backend.fetch('s3ql_metadata')
- except CorruptedObjectError:
- raise QuietError('Backend data corrupted, or file system '
- 'revision needs upgrade.', exitcode=17)
+# except CorruptedObjectError:
+# raise QuietError('Backend data corrupted, or file system '
+# 'revision needs upgrade.', exitcode=17)

return lambda: ComprencBackend(data_pw, compress, options.backend_class(options))

Best,
-Nikolaus

[Florian Ludwig]

Hi,

thanks you for the pointer. After debugging through the code I found the
culprit.... My --authfile did have an "fs-passphrase" entry which contained
a placeholder.

:]

Now everything looks good.

[d***@gmail.com]

Hi,
I have been using S3QL on OVH swift storage with OK results.. my feeling is
AWS S3 is more "solid" but OVH has been OK, for the price and I continue to
use it - though it is hard to compare as I have quite different data in
each provider.

One suggestion from OVH support that has helped in my case has been to use
the option:
* --backend-options tcp-timeout=150*

with fsck, mount etc