Skip to content

PXB-3543 [trunk] : Xtrabackup is extremely slow when performing incremenal backup for 1 Million Tables #1723

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
satya-bodapati merged 3 commits into from
Jan 19, 2026
Merged

PXB-3543 [trunk] : Xtrabackup is extremely slow when performing incremenal backup for 1 Million Tables #1723

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2133ffd
PXB-3543 : Xtrabackup 8.0 is extremely slow when performing increment…
satya-bodapati Jan 16, 2026
7e4ff8f
Merge remote-tracking branch 'fork/8.0' into 8.4
satya-bodapati Jan 19, 2026
0e515f5
Merge remote-tracking branch 'fork/8.4' into trunk
satya-bodapati Jan 19, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 13 additions & 2 deletions storage/innobase/xtrabackup/src/write_filt.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,10 +72,17 @@ static bool wf_incremental_init(xb_write_filt_ctxt_t *ctxt, char *dst_name,
(cursor->page_size / 4 + 1) * cursor->page_size + UNIV_PAGE_SIZE_MAX;
cp->delta_buf_base = static_cast<byte *>(
ut::malloc_withkey(UT_NEW_THIS_FILE_PSI_KEY, buf_size));
memset(cp->delta_buf_base, 0, buf_size);
cp->delta_buf =
static_cast<byte *>(ut_align(cp->delta_buf_base, UNIV_PAGE_SIZE_MAX));

/* OPTIMIZATION: Only zero the first page (header/index page).
Previously, we zeroed the entire buffer (~67MB), which triggered massive
OS page faults (asm_exc_page_fault) and consumed significant CPU (~80%).
We only need the header page to be clean so that no garbage follows the
0xFFFFFFFF sentinel. The data pages are overwritten by memcpy or never
written to disk.*/
memset(cp->delta_buf, 0, cursor->page_size);

/* write delta meta info */
snprintf(meta_name, sizeof(meta_name), "%s%s", dst_name,
XB_DELTA_INFO_SUFFIX);
Expand Down Expand Up @@ -135,7 +142,11 @@ static bool wf_incremental_process(xb_write_filt_ctxt_t *ctxt,
}

/* clear buffer */
memset(cp->delta_buf, 0, page_size / 4 * page_size);
/* OPTIMIZATION: Only clear the header page for the next batch.
The rest of the buffer will be overwritten by new data pages.
Since we track 'npages', we never write the dirty tail to disk,
so zeroing the whole buffer is wasted memory bandwidth.*/
memset(cp->delta_buf, 0, page_size);
/*"xtra"*/
mach_write_to_4(cp->delta_buf, 0x78747261UL);
cp->npages = 1;
Expand Down