Commit Graph

138 Commits

Author SHA1 Message Date
Girish Ramakrishnan df98847535 Remove backupDone from storage interface
This is only used by caas
2018-07-27 12:08:19 -07:00
Girish Ramakrishnan 2a18070016 do-spaces: Force retry of 4xx error codes when copying 2018-03-21 15:41:21 -07:00
Girish Ramakrishnan e0ece06b26 s3: improved copy logging 2018-03-21 14:22:41 -07:00
Girish Ramakrishnan 6adf88a6e5 Make uploads work with very slow upload speeds
chunk uploads get a timeout of 2mins (derived from http.timeout).
On servers like kimsufi, uploads takes forever (100 MB/sec limit).
Currently, our upload code does not dynamically adapt itself to
changing the concurrency when network is slow.
2018-03-20 19:37:45 -07:00
Girish Ramakrishnan ce33681c37 Dump etag info 2018-03-20 18:19:14 -07:00
Girish Ramakrishnan 565eed015f Add better backup logs 2018-03-20 16:41:45 -07:00
Girish Ramakrishnan 98e19e6df5 fix upload errors causing double callback 2018-02-27 19:16:03 -08:00
Girish Ramakrishnan b6fb49956f s3: better debug output when copying 2018-02-22 12:41:18 -08:00
Girish Ramakrishnan 3c8c15db01 s3: use a constant backoff since it takes forever to fail otherwise 2018-02-22 12:30:44 -08:00
Girish Ramakrishnan cea83889ec s3: Fix issue where it takes forever to timeout if the backend is down 2018-02-22 12:19:23 -08:00
Girish Ramakrishnan 2ecb66afd7 s3: cleanup code 2018-02-22 12:16:01 -08:00
Girish Ramakrishnan e6c07fc148 merge the done callback into the main code 2018-02-22 11:06:28 -08:00
Girish Ramakrishnan 1f30a4f3ea Make s3.deleteObjects return error 2018-02-22 11:05:29 -08:00
Girish Ramakrishnan 0bfdaeb2fb rename to chunkSize 2018-02-22 11:01:04 -08:00
Girish Ramakrishnan e022dbf8a6 Revert "merge the done callback into the main code"
This reverts commit c39bec8cc1.

This was committed with extra stuff by mistake
2018-02-22 10:58:56 -08:00
Girish Ramakrishnan 6075a7a890 typo 2018-02-22 10:34:48 -08:00
Girish Ramakrishnan 28b864c346 sos: Copy in 96M chunks as recommended by exoscale 2018-02-22 10:31:56 -08:00
Girish Ramakrishnan c39bec8cc1 merge the done callback into the main code 2018-02-21 20:17:58 -08:00
Girish Ramakrishnan 727a25f491 DO Spaces: multipart copy now works 2018-02-20 14:48:03 -08:00
Girish Ramakrishnan 77fe595970 lint 2018-02-04 00:15:41 -08:00
Girish Ramakrishnan 14529d313a sos: CopySource requires fancier encoding rules 2018-02-04 00:12:28 -08:00
Girish Ramakrishnan f607010396 sos: remove exoscale specific hacks which are obsolete now 2018-02-02 20:13:51 -08:00
Girish Ramakrishnan 568397ec19 caas: send ids in backupDone instead of filenames 2017-11-29 12:39:10 -08:00
Girish Ramakrishnan 693bc094cc caas: make fqdn part of dns and s3 credentials 2017-11-28 22:44:40 -08:00
Girish Ramakrishnan 76312495fd Add debug 2017-11-28 15:00:33 -08:00
Girish Ramakrishnan f9dda85a38 Fix error code handling 2017-11-22 21:03:34 -08:00
Girish Ramakrishnan f09c89e33f Remove confusing batchSize logic from listDir
This also fixes a bug in removeDir in DO spaces

thanks to @syn for reporting
2017-10-29 19:04:10 -07:00
Girish Ramakrishnan d127b25f0f Only set the custom https agent for HTTPS minio
Otherwise, we get a Cannot set property ‘agent’ of undefined error
2017-10-26 18:38:45 -07:00
Girish Ramakrishnan dd68c8f91f Various backup fixes 2017-10-12 11:02:48 -07:00
Girish Ramakrishnan 1444bb038f only upload needs to be retried
copy/delete are already retried in the sdk code
2017-10-11 00:08:41 -07:00
Girish Ramakrishnan ac9e421ecf improved backup progress and logging 2017-10-10 22:49:38 -07:00
Girish Ramakrishnan 78807782df Various hacks for exoscale-sos
SOS does not like multipart uploads. They just fail randomly.

As a fix, we try to detect filesystem files and skip multipart uploads
for files < 5GB. For > 5GB, we do multipart upload anyways (mostly fails).

The box backup is switched to flat-file for exoscale for the reason
above.
2017-10-10 11:03:20 -07:00
Girish Ramakrishnan 9f97f48634 Add note on s3ForcePathStyle 2017-10-09 18:46:14 -07:00
Girish Ramakrishnan 91ec2eaaf5 sos: "/" must separate bucket and key name 2017-10-09 11:50:22 -07:00
Girish Ramakrishnan f0dd90a1f5 listObjectsV2 does not work on some S3 providers
specifically, cloudscale does not support it
2017-10-05 12:07:14 -07:00
Girish Ramakrishnan ee8ee8e786 KeyCount is not set on some S3 providers 2017-10-05 11:36:54 -07:00
Girish Ramakrishnan df6e6cb071 Allow s3 backend to accept self-signed certs
Fixes #316
2017-10-05 10:14:55 -07:00
Girish Ramakrishnan ca502a2d55 Display error code 2017-10-04 22:34:44 -07:00
Girish Ramakrishnan b9efb0b50b Fix callback invokation 2017-10-04 19:28:40 -07:00
Girish Ramakrishnan c3d7d867be Do not set logCallback 2017-10-04 12:32:12 -07:00
Girish Ramakrishnan 96c16cd5d2 remove debug 2017-10-04 11:54:17 -07:00
Girish Ramakrishnan af182e3df6 caas: cache the creds, otherwise we bombard the server 2017-10-04 11:49:38 -07:00
Girish Ramakrishnan d70ff7cd5b Make copy() return event emitter
This way the storage logic does not need to rely on progress
2017-10-04 11:02:50 -07:00
Johannes Zellner 38331e71e2 Ensure all S3 CopySource properties are URI encoded 2017-10-04 19:07:08 +02:00
Johannes Zellner 322a9a18d7 Use multipart copy for s3 and files larger than 5GB 2017-10-04 18:56:23 +02:00
Girish Ramakrishnan b85bc3aa01 s3: Must encode copySource
https://github.com/aws/aws-sdk-js/issues/1302
2017-10-03 15:51:05 -07:00
Girish Ramakrishnan 01154d0ae6 s3: better error messages 2017-10-03 14:46:59 -07:00
Girish Ramakrishnan 6494050d66 Make removeDir less noisy 2017-10-03 01:22:37 -07:00
Girish Ramakrishnan 21afc71d89 add tests for storage backends 2017-10-02 23:08:16 -07:00
Girish Ramakrishnan 0f543e6703 s3: add progress detail
this is a bit of a hack and we should add another way to set the progress
(maybe via backups.setProgress or via a progress callback). this is because
some methods like removeDir can be called from backuptask and from box code.
2017-10-01 18:25:51 -07:00