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

Implement journal write batching and slightly refactor journal writes

Browse Source 
Slightly reduces WA. For example, in 4K T1Q128 replicated randwrite tests
WA is reduced from ~3.6 to ~3.1, in T1Q64 from ~3.8 to ~3.4.

Only effective without no_same_sector_overwrites.
This commit is contained in:
Vitaliy Filippov
2021-12-16 00:27:17 +03:00
parent 999bed8514
commit f93491bc6c
8 changed files with 163 additions and 219 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/blockstore_journal.cpp
+65 -14
View File
@@ -153,22 +153,73 @@ journal_entry* prefill_single_journal_entry(journal_t & journal, uint16_t type,
return je;
}
void prepare_journal_sector_write(journal_t & journal, int cur_sector, io_uring_sqe *sqe, std::function<void(ring_data_t*)> cb)
void blockstore_impl_t::prepare_journal_sector_write(int cur_sector, blockstore_op_t *op)
{
// Don't submit the same sector twice in the same batch
if (!journal.sector_info[cur_sector].submit_id)
{
io_uring_sqe *sqe = get_sqe();
// Caller must ensure availability of an SQE
assert(sqe != NULL);
ring_data_t *data = ((ring_data_t*)sqe->user_data);
journal.sector_info[cur_sector].written = true;
journal.sector_info[cur_sector].submit_id = ++journal.submit_id;
journal.submitting_sectors.push_back(cur_sector);
journal.sector_info[cur_sector].flush_count++;
data->iov = (struct iovec){
(journal.inmemory
? journal.buffer + journal.sector_info[cur_sector].offset
: journal.sector_buf + journal.block_size*cur_sector),
journal.block_size
};
data->callback = [this, flush_id = journal.submit_id](ring_data_t *data) { handle_journal_write(data, flush_id); };
my_uring_prep_writev(
sqe, journal.fd, &data->iov, 1, journal.offset + journal.sector_info[cur_sector].offset
);
}
journal.sector_info[cur_sector].dirty = false;
journal.sector_info[cur_sector].written = true;
journal.sector_info[cur_sector].flush_count++;
ring_data_t *data = ((ring_data_t*)sqe->user_data);
data->iov = (struct iovec){
(journal.inmemory
? journal.buffer + journal.sector_info[cur_sector].offset
: journal.sector_buf + journal.block_size*cur_sector),
journal.block_size
};
data->callback = cb;
my_uring_prep_writev(
sqe, journal.fd, &data->iov, 1, journal.offset + journal.sector_info[cur_sector].offset
);
// But always remember that this operation has to wait until this exact journal write is finished
journal.flushing_ops.insert((pending_journaling_t){
.flush_id = journal.sector_info[cur_sector].submit_id,
.sector = cur_sector,
.op = op,
});
auto priv = PRIV(op);
priv->pending_ops++;
if (!priv->min_flushed_journal_sector)
priv->min_flushed_journal_sector = 1+cur_sector;
priv->max_flushed_journal_sector = 1+cur_sector;
}
void blockstore_impl_t::handle_journal_write(ring_data_t *data, uint64_t flush_id)
{
live = true;
if (data->res != data->iov.iov_len)
{
// FIXME: our state becomes corrupted after a write error. maybe do something better than just die
throw std::runtime_error(
"journal write failed ("+std::to_string(data->res)+" != "+std::to_string(data->iov.iov_len)+
"). in-memory state is corrupted. AAAAAAAaaaaaaaaa!!!111"
);
}
auto fl_it = journal.flushing_ops.upper_bound((pending_journaling_t){ .flush_id = flush_id });
if (fl_it != journal.flushing_ops.end() && fl_it->flush_id == flush_id)
{
journal.sector_info[fl_it->sector].flush_count--;
}
while (fl_it != journal.flushing_ops.end() && fl_it->flush_id == flush_id)
{
auto priv = PRIV(fl_it->op);
priv->pending_ops--;
assert(priv->pending_ops >= 0);
if (priv->pending_ops == 0)
{
release_journal_sectors(fl_it->op);
priv->op_state++;
ringloop->wakeup();
}
journal.flushing_ops.erase(fl_it++);
}
}
journal_t::~journal_t()