GGLinnk/bees
1
0
Fork 0
mirror of https://github.com/Zygo/bees.git synced 2025-06-17 01:56:16 +02:00
Code Packages Activity

hash: don't spin when writes fail

Browse Source 
When a hash table write fails, we skip over the write throttling because
we didn't report that we successfully wrote an extent.  This can be bad
if the filesystem is full and the allocations for writes are burning a
lot of CPU time searching for free space.

We also don't retry the write later on since we assume the extent is
clean after a write attempt whether it was successful or not, so the
extent might not be written out later when writes are possible again.

Check whether a hash extent is dirty, and always throttle after
attempting the write.

If a write fails, leave the extent dirty so we attempt to write it out
the next time flush cycles through the hash table.  During shutdown
this will reattempt each failing write once, after that the updated hash
table data will be dropped.

Signed-off-by: Zygo Blaxell <bees@furryterror.org>
This commit is contained in:
Zygo Blaxell
2023-01-19 09:24:48 -05:00
parent 28ee2ae1a8
commit cbc76a7457
2 changed files with 12 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
src/bees-hash.cc
View File

@ -106,12 +106,6 @@ BeesHashTable::flush_dirty_extent(uint64_t extent_index)
BEESNOTE("flushing extent #" << extent_index << " of " << m_extents << " extents");
auto lock = lock_extent_by_index(extent_index);
// Not dirty, nothing to do
if (!m_extent_metadata.at(extent_index).m_dirty) {
return false;
}
bool wrote_extent = false;
catch_all([&]() {
@ -125,9 +119,6 @@ BeesHashTable::flush_dirty_extent(uint64_t extent_index)
// Copy the extent because we might be stuck writing for a while
ByteVector extent_copy(dirty_extent, dirty_extent_end);
// Mark extent non-dirty while we still hold the lock
m_extent_metadata.at(extent_index).m_dirty = false;
// Release the lock
lock.unlock();
@ -139,6 +130,10 @@ BeesHashTable::flush_dirty_extent(uint64_t extent_index)
// const size_t dirty_extent_size = dirty_extent_end - dirty_extent;
// bees_unreadahead(m_fd, dirty_extent_offset, dirty_extent_size);
// Mark extent clean if write was successful
lock.lock();
m_extent_metadata.at(extent_index).m_dirty = false;
wrote_extent = true;
});
@ -152,6 +147,13 @@ BeesHashTable::flush_dirty_extents(bool slowly)
uint64_t wrote_extents = 0;
for (size_t extent_index = 0; extent_index < m_extents; ++extent_index) {
// Skip the clean ones
auto lock = lock_extent_by_index(extent_index);
if (!m_extent_metadata.at(extent_index).m_dirty) {
continue;
}
lock.unlock();
if (flush_dirty_extent(extent_index)) {
++wrote_extents;
if (slowly) {