thebestvibedev/tromcho.net
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix OSDs still crashing when the cluster is full with EC

Browse Source 
ENOSPC handling was introduced in 1.6.0 but it was not complete; now it is

P.S: See also client_retry_enospc (true by default)
This commit is contained in:
Vitaliy Filippov
2024-12-26 01:56:33 +03:00
parent cef98052f5
commit 42eebfc1bd
2 changed files with 10 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/osd/osd_primary_subops.cpp
+2 -2
View File
@@ -744,7 +744,7 @@ void osd_t::submit_primary_rollback_subops(osd_op_t *cur_op, const uint64_t* osd
for (int role = 0; role < op_data->pg_size; role++) for (int role = 0; role < op_data->pg_size; role++)
{ {
if (osd_set[role] != 0 && !stripes[role].read_error && if (osd_set[role] != 0 && !stripes[role].read_error &&
msgr.osd_peer_fds.find(osd_set[role]) != msgr.osd_peer_fds.end()) (osd_set[role] == this->osd_num || msgr.osd_peer_fds.find(osd_set[role]) != msgr.osd_peer_fds.end()))
{ {
n_subops++; n_subops++;
} }
@@ -761,7 +761,7 @@ void osd_t::submit_primary_rollback_subops(osd_op_t *cur_op, const uint64_t* osd
for (int role = 0; role < op_data->pg_size; role++) for (int role = 0; role < op_data->pg_size; role++)
{ {
if (osd_set[role] != 0 && !stripes[role].read_error && if (osd_set[role] != 0 && !stripes[role].read_error &&
msgr.osd_peer_fds.find(osd_set[role]) != msgr.osd_peer_fds.end()) (osd_set[role] == this->osd_num || msgr.osd_peer_fds.find(osd_set[role]) != msgr.osd_peer_fds.end()))
{ {
osd_op_t *subop = &op_data->subops[i]; osd_op_t *subop = &op_data->subops[i];
op_data->unstable_writes[i] = (obj_ver_id){ op_data->unstable_writes[i] = (obj_ver_id){
src/osd/osd_primary_write.cpp
+8 -6
View File
@@ -260,11 +260,6 @@ resume_4:
op_data->st = 4; op_data->st = 4;
return; return;
resume_5: resume_5:
if (op_data->scheme != POOL_SCHEME_REPLICATED)
{
// Remove version override just after the write, but before stabilizing
pg.ver_override.erase(op_data->oid);
}
if (op_data->errors > 0) if (op_data->errors > 0)
{ {
// Handle ENOSPC/EDOM/ERANGE/EIO. If some subops fail, but others succeed, // Handle ENOSPC/EDOM/ERANGE/EIO. If some subops fail, but others succeed,
@@ -276,7 +271,7 @@ resume_5:
{ {
if (op_data->scheme != POOL_SCHEME_REPLICATED) if (op_data->scheme != POOL_SCHEME_REPLICATED)
{ {
submit_primary_rollback_subops(cur_op, op_data->prev_set); submit_primary_rollback_subops(cur_op, pg.cur_set.data());
resume_11: resume_11:
op_data->st = 11; op_data->st = 11;
return; return;
@@ -287,15 +282,22 @@ resume_12:
} }
else else
{ {
pg.ver_override.erase(op_data->oid);
mark_partial_write(pg, op_data->oid, op_data->object_state, op_data->stripes, true); mark_partial_write(pg, op_data->oid, op_data->object_state, op_data->stripes, true);
pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode); pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode);
return; return;
} }
} }
pg.ver_override.erase(op_data->oid);
deref_object_state(pg, &op_data->object_state, true); deref_object_state(pg, &op_data->object_state, true);
pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode); pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode);
return; return;
} }
if (op_data->scheme != POOL_SCHEME_REPLICATED)
{
// Remove version override just after the write, but before stabilizing
pg.ver_override.erase(op_data->oid);
}
if (op_data->object_state) if (op_data->object_state)
{ {
// We must forget the unclean state of the object before deleting it // We must forget the unclean state of the object before deleting it