mirror of
https://github.com/matrix-construct/construct
synced 2024-12-27 07:54:05 +01:00
ircd::db: Remove offload shenanigans.
This commit is contained in:
Jason Volk
parent
31ff4f68cb
commit
850fd0238a
1 changed files with 7 additions and 190 deletions
191
ircd/db.cc
191
ircd/db.cc
|
|
@ -5620,7 +5620,6 @@ namespace ircd::db
|
|||
static rocksdb::Iterator &_seek_(rocksdb::Iterator &, const string_view &);
|
||||
static rocksdb::Iterator &_seek_lower_(rocksdb::Iterator &, const string_view &);
|
||||
static rocksdb::Iterator &_seek_upper_(rocksdb::Iterator &, const string_view &);
|
||||
static std::unique_ptr<rocksdb::Iterator> _seek_offload(database::column &c, const rocksdb::ReadOptions &opts, const std::function<void (rocksdb::Iterator &)> &closure);
|
||||
bool _seek(database::column &, const pos &, const rocksdb::ReadOptions &, std::unique_ptr<rocksdb::Iterator> &it);
|
||||
bool _seek(database::column &, const string_view &, const rocksdb::ReadOptions &, std::unique_ptr<rocksdb::Iterator> &it);
|
||||
}
|
||||
|
|
@ -5660,8 +5659,6 @@ ircd::db::seek(database::column &c,
|
|||
const rocksdb::ReadOptions &opts,
|
||||
std::unique_ptr<rocksdb::Iterator> &it)
|
||||
{
|
||||
assert(opts.read_tier == NON_BLOCKING);
|
||||
|
||||
if(!it)
|
||||
{
|
||||
database &d(*c.d);
|
||||
|
|
@ -5672,12 +5669,6 @@ ircd::db::seek(database::column &c,
|
|||
return _seek(c, p, opts, it);
|
||||
}
|
||||
|
||||
//
|
||||
// Seek with offload-safety in case of blocking IO.
|
||||
//
|
||||
// The options for an iterator cannot be changed after the iterator is created.
|
||||
// This slightly complicates our toggling between blocking and non-blocking queries.
|
||||
//
|
||||
bool
|
||||
ircd::db::_seek(database::column &c,
|
||||
const string_view &p,
|
||||
|
|
@ -5687,82 +5678,24 @@ ircd::db::_seek(database::column &c,
|
|||
database &d(*c.d);
|
||||
const ircd::timer timer;
|
||||
|
||||
// Start with a non-blocking query.
|
||||
_seek_(*it, p);
|
||||
|
||||
// Branch for query being fulfilled from cache
|
||||
if(!it->status().IsIncomplete())
|
||||
{
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK %s %s in %ld$us '%s'",
|
||||
log, "'%s' %lu:%lu SEEK %s in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
it->status().ToString(),
|
||||
valid(*it)? "VALID" : "INVALID",
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return valid(*it);
|
||||
}
|
||||
|
||||
const auto blocking_it
|
||||
{
|
||||
_seek_offload(c, opts, [&p](rocksdb::Iterator &blocking_it)
|
||||
{
|
||||
_seek_(blocking_it, p);
|
||||
})
|
||||
};
|
||||
|
||||
// When the blocking iterator comes back invalid the result is propagated
|
||||
if(!valid(*blocking_it))
|
||||
{
|
||||
it.reset(rocksdb::NewErrorIterator(blocking_it->status()));
|
||||
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK %s INVALID CACHE MISS in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
it->status().ToString(),
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
it.reset(nullptr);
|
||||
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK %s VALID CACHE MISS in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
blocking_it->status().ToString(),
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return seek(c, slice(blocking_it->key()), opts, it);
|
||||
}
|
||||
|
||||
//
|
||||
// Seek with offload-safety in case of blocking IO.
|
||||
//
|
||||
// The options for an iterator cannot be changed after the iterator is created.
|
||||
// This slightly complicates our toggling between blocking and non-blocking queries.
|
||||
//
|
||||
bool
|
||||
ircd::db::_seek(database::column &c,
|
||||
const pos &p,
|
||||
|
|
@ -5776,140 +5709,24 @@ ircd::db::_seek(database::column &c,
|
|||
valid(*it)
|
||||
};
|
||||
|
||||
const size_t it_key_size
|
||||
{
|
||||
valid_it? size(slice(it->key())) : 0
|
||||
};
|
||||
|
||||
if(unlikely(it_key_size > 1024))
|
||||
throw error
|
||||
{
|
||||
"Iteration with key size %zu not supported atm", it_key_size
|
||||
};
|
||||
|
||||
char it_key_buf[it_key_size];
|
||||
const mutable_buffer it_key_mb
|
||||
{
|
||||
it_key_buf, it_key_size
|
||||
};
|
||||
|
||||
// Unfortunately we have to copy the key unless we use a pinned iterator.
|
||||
// In the future this offloading system should disappear when the env impl
|
||||
// is completed to use AIO. Right now the majority case for this copy is
|
||||
// only an 8 byte key...
|
||||
const auto it_key
|
||||
{
|
||||
valid_it?
|
||||
string_view{it_key_buf, copy(it_key_mb, slice(it->key()))}:
|
||||
string_view{}
|
||||
};
|
||||
|
||||
// Start with a non-blocking query. This used to not invalidate the
|
||||
// iterator when it failed with a cache miss (incomplete), but now it
|
||||
// does (hence the copy above).
|
||||
_seek_(*it, p);
|
||||
|
||||
// Branch for query being fulfilled from cache
|
||||
if(!it->status().IsIncomplete())
|
||||
{
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK[%s] %s %s -> %s in %ld$us '%s'",
|
||||
log, "'%s' %lu:%lu SEEK[%s] %s -> %s in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
reflect(p),
|
||||
it->status().ToString(),
|
||||
valid_it? "VALID" : "INVALID",
|
||||
valid(*it)? "VALID" : "INVALID",
|
||||
it->status().ToString(),
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return valid(*it);
|
||||
}
|
||||
|
||||
const auto blocking_it
|
||||
{
|
||||
_seek_offload(c, opts, [&valid_it, &it_key, &p]
|
||||
(rocksdb::Iterator &blocking_it)
|
||||
{
|
||||
if(valid_it)
|
||||
_seek_(blocking_it, it_key);
|
||||
|
||||
_seek_(blocking_it, p);
|
||||
})
|
||||
};
|
||||
|
||||
// When the blocking iterator comes back invalid the result is propagated
|
||||
if(!valid(*blocking_it))
|
||||
{
|
||||
it.reset(rocksdb::NewErrorIterator(blocking_it->status()));
|
||||
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK[%s] %s %s -> %s|INVALID CACHE MISS in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
reflect(p),
|
||||
it->status().ToString(),
|
||||
valid_it? "VALID" : "INVALID",
|
||||
valid(*it)? "VALID" : "INVALID",
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
it.reset(nullptr);
|
||||
|
||||
#ifdef RB_DEBUG_DB_SEEK
|
||||
log::debug
|
||||
{
|
||||
log, "'%s' %lu:%lu SEEK[%s] %s %s -> VALID CACHE MISS in %ld$us '%s'",
|
||||
name(d),
|
||||
sequence(d),
|
||||
sequence(opts.snapshot),
|
||||
reflect(p),
|
||||
blocking_it->status().ToString(),
|
||||
valid_it? "VALID" : "INVALID",
|
||||
timer.at<microseconds>().count(),
|
||||
name(c)
|
||||
};
|
||||
#endif
|
||||
|
||||
return seek(c, slice(blocking_it->key()), opts, it);
|
||||
}
|
||||
|
||||
/// DB cache miss: create a blocking iterator and offload it.
|
||||
std::unique_ptr<rocksdb::Iterator>
|
||||
ircd::db::_seek_offload(database::column &c,
|
||||
const rocksdb::ReadOptions &opts,
|
||||
const std::function<void (rocksdb::Iterator &)> &closure)
|
||||
{
|
||||
database &d(*c.d);
|
||||
rocksdb::ColumnFamilyHandle *const &cf(c);
|
||||
rocksdb::ReadOptions blocking_opts(opts);
|
||||
blocking_opts.fill_cache = true;
|
||||
blocking_opts.read_tier = BLOCKING;
|
||||
std::unique_ptr<rocksdb::Iterator> blocking_it
|
||||
{
|
||||
d.d->NewIterator(blocking_opts, cf)
|
||||
};
|
||||
|
||||
const auto function{[&closure, &blocking_it]
|
||||
{
|
||||
closure(*blocking_it);
|
||||
}};
|
||||
|
||||
ctx::offload(function);
|
||||
return blocking_it;
|
||||
}
|
||||
|
||||
/// Seek to entry NOT GREATER THAN key. That is, equal to or less than key
|
||||
|
|
@ -6336,7 +6153,7 @@ ircd::db::make_opts(const gopts &opts)
|
|||
{
|
||||
rocksdb::ReadOptions ret;
|
||||
assert(ret.fill_cache);
|
||||
ret.read_tier = NON_BLOCKING;
|
||||
ret.read_tier = BLOCKING;
|
||||
|
||||
// slice* for exclusive upper bound. when prefixes are used this value must
|
||||
// have the same prefix because ordering is not guaranteed between prefixes
|
||||
|
|
|
|||
Loading…
Reference in a new issue