Commit Graph

487 Commits (824ea507d0ec15db51d46b08c0927501fd760f6a)

Author SHA1 Message Date
Vitaliy Filippov 824ea507d0 Do not try to push more segments than IOV_MAX at once as it leads to EMSGSIZE 2020-10-30 01:25:43 +03:00
Vitaliy Filippov 23ea409081 Fix "can't get SQE, will fall out of sync with EPOLLET" when overflowing the ring
OSDs shouldn't crash or hang with long iodepths anymore
2020-10-30 01:06:36 +03:00
Vitaliy Filippov 2ccb75974b Fix a rare crash caused by a stopped client still being in write_ready_clients 2020-10-30 01:04:58 +03:00
Vitaliy Filippov 6561d4e040 Validate pool ID before executing the operation
Reply -EPIPE for non-existing pools because we assume that it means
that pool config isn't loaded yet. Previously OSD crashed on such operations
2020-10-30 01:02:46 +03:00
Vitaliy Filippov 1eda7f529d Note about Linux 5.8+ 2020-10-28 19:17:22 +03:00
Vitaliy Filippov 0a174bb313 Return success for already finished rollback operations
There was a FIXME and I actually hit it during tests :)
2020-10-24 18:46:19 +03:00
Vitaliy Filippov 720985e4c7 Fix NULL rmw buffer after the latest changes and add a testcase for it 2020-10-24 18:29:19 +03:00
Vitaliy Filippov 4872f617a4 Clear connect timeout in stop_client() to stop races during disconnections 2020-10-24 10:37:16 +03:00
Vitaliy Filippov e8ac08be14 Allow to overwrite incomplete objects or parts of objects to recover them 2020-10-24 02:14:41 +03:00
Vitaliy Filippov 660c2412fb Improve debugging output for incomplete/degraded 2020-10-24 01:28:47 +03:00
Vitaliy Filippov faa5e1436f Attempt journal trim even without new flushes
This is the new guaranteed unblocking method which replaces old trims
in init and rollback, and also fixes a possible stall when just several
writes in the beginning of the journal are flushed without triggering
a subsequent trim.
2020-10-24 01:28:47 +03:00
Vitaliy Filippov 5fbe36198a Fix journal trimming
1) Update journal's used_start in memory only after updating journal superblock.
Doing the opposite is incorrect because part of the journal will be lost if writers
overwrite its old beginning.

2) Sync journal device after updating the superblock.

3) Do not trim in rollback and init because trimming there would also require
updating the superblock. And the only reason to trim in both those places was
to unblock writers. And a guaranteed unblocking method will follow in the next
commit :)
2020-10-24 01:08:33 +03:00
Vitaliy Filippov 99c45bb5ed Fix debugging output during journal loading 2020-10-24 01:08:33 +03:00
Vitaliy Filippov 701eb79422 Stabilize writes before deleting extra chunks to not stall peer journals 2020-10-23 22:45:05 +03:00
Vitaliy Filippov 220bda0667 Fix possible buffer over(under)flow when handling LIST 2020-10-23 02:17:44 +03:00
Vitaliy Filippov 1e8f0328e0 Cancel outbound operations after re-peering PGs 2020-10-22 22:54:38 +00:00
Vitaliy Filippov f011e0c675 Do not block stabilize by list and list by write 2020-10-22 22:13:40 +00:00
Vitaliy Filippov 1a694c387e Print slow ops in log 2020-10-20 23:41:23 +00:00
Vitaliy Filippov 738ad5af79 Fix infinite looping in continue_recovery_op() when pg_cancel_write_queue() is called 2020-10-20 22:23:15 +00:00
Vitaliy Filippov 9abf3c17c9 Correct fix for "Pool %u PG %u configuration is invalid" during startup
Establish watcher connection after loading PGs
2020-10-20 21:09:14 +00:00
Vitaliy Filippov d2b901aa09 Fix default auto-created failure domains 2020-10-20 21:07:40 +00:00
Vitaliy Filippov befff09370 Fix possible crash due to uninitialized ring_data_t in ringloop 2020-10-20 10:44:38 +03:00
Vitaliy Filippov d1645551d4 Implement write batching
Also fix possible race condition which could in theory lead to "command out of sync"
and a buffer overflow that could happen on incorrect server response.
2020-10-20 03:29:17 +03:00
Vitaliy Filippov 7cb561f95a Add etcd to the example service generator 2020-10-20 01:50:56 +03:00
Vitaliy Filippov ae480196e2 Add a note about etcd bug, fix simple-offsets.js cmdline 2020-10-19 17:05:45 +03:00
Vitaliy Filippov 398c86f943 Improve PG-related log messages 2020-10-18 12:17:22 +00:00
Vitaliy Filippov bec5f921a6 Fix buffer overflows in the no_same_sector_overwrites mode 2020-10-17 23:30:16 +00:00
Vitaliy Filippov 5335c8de8e Do not use unordered_map for list_ops/list_results 2020-10-17 23:30:16 +00:00
Vitaliy Filippov c696a82083 Replace assert with if + error message (may happen on metadata corruption) 2020-10-17 23:30:16 +00:00
Vitaliy Filippov 900171586b XOR 2+1 test results 2020-10-17 14:58:08 +03:00
Vitaliy Filippov 70612e5df0 Do not handle change events before loading config 2020-10-17 11:18:39 +00:00
Vitaliy Filippov d952c24979 Use timeout in rw callback 2020-10-17 11:00:55 +00:00
Vitaliy Filippov 776fe954a5 Fix crashes on multiple OSD reconnects
Identify clients by pointers instead of peer_fd as peer may be dropped
and reconnected between callbacks

Yeah maybe I need some Rust, but ... maybe in the future :)
2020-10-17 10:53:04 +00:00
Vitaliy Filippov 9350656af6 Fix osd tags 2020-10-16 23:28:48 +00:00
Vitaliy Filippov ece14a7d65 Hide "Connected with..." client messages by default 2020-10-11 02:22:46 +03:00
Vitaliy Filippov be5f314c32 Change notes about gcc requirement to 9+, fio to 3.16+ 2020-10-11 02:00:39 +03:00
Vitaliy Filippov 15dba96375 Implement inode removal tool. Removes multiple objects from multiple OSDs in parallel 2020-10-10 01:08:19 +03:00
Vitaliy Filippov 3d05aa9362 Make it build with GCC 10, fio 3.20+ (atomics...) and QEMU 5.1 2020-10-06 02:35:11 +03:00
Vitaliy Filippov 94efb54feb Implement OSD tags (device classes), fix pool failure_domain configuration 2020-10-04 17:31:50 +03:00
Vitaliy Filippov aa2a0ee00f Do not group adjacent stripes by default as it's pointless on SSDs 2020-10-02 10:17:54 +03:00
Vitaliy Filippov 9f2a948712 Make pg_stripe_size a per-pool config 2020-10-01 18:51:49 +03:00
Vitaliy Filippov ba74eece4a More fixes to the failure model (why am I doing this?..) 2020-10-01 18:38:30 +03:00
Vitaliy Filippov 2fdd8a1b38 More correct failure model (I hope so) 2020-10-01 02:33:48 +03:00
Vitaliy Filippov 526983f7a9 Add usable CLI commands for NBD proxy (map/unmap/list) 2020-09-29 02:06:19 +03:00
Vitaliy Filippov 8e36f04482 One more experiment with cluster AFR% 2020-09-27 19:42:42 +03:00
Vitaliy Filippov f460d8c1c8 Add note about NBD 2020-09-26 00:11:55 +03:00
Vitaliy Filippov 7619a789c0 Set request size in NBD 2020-09-26 00:01:23 +03:00
Vitaliy Filippov e65a28e27e Implement a simple NBD proxy (does not daemonize yet) 2020-09-25 20:51:01 +03:00
Vitaliy Filippov 6852f299ae Add functions to calculate AFR for a cluster 2020-09-24 23:15:26 +03:00
Vitaliy Filippov 1967269c13 Resume operations in cluster_client when PGs are loaded (fixes a hang in qemu-img) 2020-09-20 01:50:19 +03:00