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

Return listing errors from list_inode_start(), abort merging and fail deletion on unsuccessfull listings

Browse Source
This commit is contained in:
Vitaliy Filippov
2024-12-27 18:31:21 +03:00
parent 5c890e4a12
commit ba9aabf187
4 changed files with 65 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/client/cluster_client.h
+2 -1
View File
@@ -136,8 +136,9 @@ public:
bool get_immediate_commit(uint64_t inode);
void continue_ops(int time_passed = 0);
// FIXME: list_inode_start/list_inode_next - not an ideal interface :)
inode_list_t *list_inode_start(inode_t inode,
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback);
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int errcode, int status)> callback);
int list_pg_count(inode_list_t *lst);
const std::vector<osd_num_t> & list_inode_get_inactive_osds(inode_list_t *lst);
const std::vector<pg_num_t> & list_inode_get_inactive_pgs(inode_list_t *lst);
src/client/cluster_client_list.cpp
+6 -3
View File
@@ -20,6 +20,7 @@ struct inode_list_pg_t
{
inode_list_t *lst = NULL;
int pos = 0;
int errcode = 0;
pg_num_t pg_num;
osd_num_t cur_primary;
bool has_unstable = false;
@@ -40,11 +41,11 @@ struct inode_list_t
std::vector<osd_num_t> inactive_osds;
std::vector<pg_num_t> inactive_pgs;
std::vector<inode_list_pg_t*> pgs;
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback;
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int errcode, int status)> callback;
};
inode_list_t* cluster_client_t::list_inode_start(inode_t inode,
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback)
std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int errcode, int status)> callback)
{
init_msgr();
pool_id_t pool_id = INODE_POOL(inode);
@@ -211,6 +212,7 @@ void cluster_client_t::continue_listing(inode_list_t *lst)
}
else if (!lst->pgs[i]->list_osds.size())
{
lst->pgs[i]->errcode = -EIO;
finish_list_pg(lst->pgs[i]);
if (check_finish_listing(lst))
{
@@ -259,6 +261,7 @@ void cluster_client_t::send_list(inode_list_osd_t *cur_list)
{
fprintf(stderr, "Failed to get PG %u/%u object list from OSD %ju (retval=%jd), skipping\n",
cur_list->pg->lst->pool_id, cur_list->pg->pg_num, cur_list->osd_num, op->reply.hdr.retval);
cur_list->pg->errcode = op->reply.hdr.retval;
}
else
{
@@ -317,7 +320,7 @@ void cluster_client_t::finish_list_pg(inode_list_pg_t *pg)
status |= INODE_LIST_HAS_UNSTABLE;
}
lst->pgs[pg->pos] = NULL;
lst->callback(lst, std::move(pg->objects), pg->pg_num, pg->cur_primary, status);
lst->callback(lst, std::move(pg->objects), pg->pg_num, pg->cur_primary, pg->errcode, status);
delete pg;
}
else
src/cmd/cli_merge.cpp
+35 -4
View File
@@ -51,6 +51,7 @@ struct snap_merger_t
btree::safe_btree_set<uint64_t> merge_offsets;
btree::safe_btree_set<uint64_t>::iterator oit;
std::map<inode_t, std::vector<uint64_t>> layer_lists;
std::map<inode_t, int> list_errcode;
std::map<inode_t, uint64_t> layer_block_size;
std::map<inode_t, uint64_t> layer_list_pos;
std::vector<snap_rw_op_t*> continue_rwo, continue_rwo2;
@@ -251,6 +252,7 @@ struct snap_merger_t
// Get parents and so on
start_merge();
// First list lower layers
list_errcode.clear();
list_layers(true);
state = 1;
resume_1:
@@ -259,6 +261,15 @@ struct snap_merger_t
// Wait for lists
return;
}
if (list_errcode.size())
{
result = (cli_result_t){
.err = EIO,
.text = "Failed to list lower layer(s) in some PGs, merging would be incorrect",
};
state = 100;
return;
}
if (merge_offsets.size() > 0)
{
state = 2;
@@ -295,6 +306,7 @@ struct snap_merger_t
state = 3;
resume_3:
// Then list upper layers
list_errcode.clear();
list_layers(false);
state = 4;
resume_4:
@@ -303,6 +315,15 @@ struct snap_merger_t
// Wait for lists
return;
}
if (list_errcode.size() > 0)
{
result = (cli_result_t){
.err = EIO,
.text = "Failed to list upper layer(s) in some PGs, merging would be incorrect",
};
state = 100;
return;
}
state = 5;
processed = 0;
to_process = merge_offsets.size();
@@ -369,8 +390,12 @@ struct snap_merger_t
{
lists_todo++;
inode_list_t* lst = parent->cli->list_inode_start(src, [this, src](
inode_list_t *lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)
inode_list_t *lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int errcode, int status)
{
if (errcode)
{
list_errcode[src] = errcode;
}
uint64_t layer_block = layer_block_size.at(src);
for (object_id obj: objects)
{
@@ -394,9 +419,15 @@ struct snap_merger_t
if (status & INODE_LIST_DONE)
{
auto & name = parent->cli->st_cli.inode_config.at(src).name;
if (parent->progress)
if (list_errcode.find(src) != list_errcode.end())
{
printf("Got listing of layer %s (inode %ju in pool %u)\n", name.c_str(), INODE_NO_POOL(src), INODE_POOL(src));
fprintf(stderr, "Failed to get listing of layer %s (inode %ju in pool %u): %s (code %d)\n",
name.c_str(), INODE_NO_POOL(src), INODE_POOL(src), strerror(-list_errcode[src]), list_errcode[src]);
}
else if (parent->progress)
{
fprintf(stderr, "Got listing of layer %s (inode %ju in pool %u)\n",
name.c_str(), INODE_NO_POOL(src), INODE_POOL(src));
}
if (delete_source)
{
@@ -428,7 +459,7 @@ struct snap_merger_t
{
if (op->retval < 0)
{
fprintf(stderr, "error reading target bitmap at offset %jx: %s\n", op->offset, strerror(-op->retval));
fprintf(stderr, "Warning: failed to read target bitmap at offset %jx: %s\n", op->offset, strerror(-op->retval));
}
else
{
src/cmd/cli_rm_data.cpp
+22 -14
View File
@@ -32,7 +32,7 @@ struct rm_inode_t
inode_list_t *lister = NULL;
std::vector<rm_pg_t*> lists;
std::vector<osd_num_t> inactive_osds;
std::vector<uint64_t> inactive_pgs;
std::set<pg_num_t> inactive_pgs;
uint64_t total_count = 0, total_done = 0, total_prev_pct = 0;
uint64_t pgs_to_list = 0;
bool lists_done = false;
@@ -44,8 +44,12 @@ struct rm_inode_t
void start_delete()
{
lister = parent->cli->list_inode_start(inode, [this](inode_list_t *lst,
std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)
std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int errcode, int status)
{
if (errcode)
{
inactive_pgs.insert(pg_num);
}
rm_pg_t *rm = new rm_pg_t((rm_pg_t){
.pg_num = pg_num,
.rm_osd_num = primary_osd,
@@ -102,17 +106,6 @@ struct rm_inode_t
}
fprintf(stderr, "\n");
}
auto inactive_pgs = parent->cli->list_inode_get_inactive_pgs(lister);
if (inactive_pgs.size() && !parent->json_output)
{
fprintf(stderr, "Some data may remain after delete in PGs which are currently inactive: ");
for (int i = 0; i < inactive_pgs.size(); i++)
{
this->inactive_pgs.push_back((uint64_t)inactive_pgs[i]);
fprintf(stderr, i > 0 ? ", %u" : "%u", inactive_pgs[i]);
}
fprintf(stderr, "\n");
}
pgs_to_list = parent->cli->list_pg_count(lister);
parent->cli->list_inode_next(lister, parent->parallel_osds);
}
@@ -237,6 +230,16 @@ struct rm_inode_t
{
fprintf(stderr, "\n");
}
if (inactive_pgs.size() && !parent->json_output)
{
fprintf(stderr, "Failed to list some PGs, deletion is not complete: PG ");
int i = 0;
for (auto pg_num: inactive_pgs)
{
fprintf(stderr, (i++) > 0 ? ", %u" : "%u", pg_num);
}
fprintf(stderr, "\n");
}
bool is_error = (total_done < total_count || inactive_osds.size() > 0 || inactive_pgs.size() > 0 || error_count > 0);
if (parent->progress && is_error)
{
@@ -246,6 +249,11 @@ struct rm_inode_t
pool_id, INODE_NO_POOL(inode), pool_id, INODE_NO_POOL(inode)
);
}
json11::Json::array inactive_pgs_json;
for (auto pg_num: inactive_pgs)
{
inactive_pgs_json.push_back((uint64_t)pg_num);
}
result = (cli_result_t){
.err = is_error && !down_ok ? EIO : 0,
.text = is_error ? "Some blocks were not removed" : (
@@ -255,7 +263,7 @@ struct rm_inode_t
{ "removed_objects", total_done },
{ "total_objects", total_count },
{ "inactive_osds", inactive_osds },
{ "inactive_pgs", inactive_pgs },
{ "inactive_pgs", inactive_pgs_json },
},
};
state = 100;