From: Chris Murphy <lists@colorremedies.com>
To: Chris Murphy <lists@colorremedies.com>
Cc: Nikolay Borisov <nborisov@suse.com>,
Btrfs BTRFS <linux-btrfs@vger.kernel.org>
Subject: Re: 5.6, slow send/receive, < 1MB/s
Date: Tue, 5 May 2020 23:48:01 -0600 [thread overview]
Message-ID: <CAJCQCtQu4ffJYuOUWkhV_wR7L0ya7mTyt0tuLqbko-O8S+1fmg@mail.gmail.com> (raw)
In-Reply-To: <CAJCQCtSCzD-RtGH1tJjNN=PBgUfJARy0r1p1Ln0pU1eRNTmR9w@mail.gmail.com>
btrfs-progs-5.6-1.fc32.x86_64
kernel-5.6.8-300.fc32.x86_64
I've created a new file system, filled it with ~600G, set the seed
flag, and I'm creating a sprout by adding a 2nd device and removing
the 1st. This used to go at device speeds. Today it's running at
~35M/s where I expect 95-120M/s. I don't know that it's related to
send/receive performance.
The remove process with 'perf top -g -U' is only using 7% CPU, unlike
the send/receive cases.
Samples
Children Self Shared Objec Symbol
- 82.39% 0.04% [kernel] [k] relocate_block_group
- 17.88% relocate_block_group
- 7.78% prepare_to_relocate
- 7.21% btrfs_commit_transaction
- 7.51% btrfs_run_delayed_refs
- __btrfs_run_delayed_refs
- 3.19% __tracepoint_kvm_mmu_spte_requested
- 3.15% pipapo_get.constprop.0
- 3.57% __add_to_free_space_tree
3.61% btrfs_del_items
1.37% btrfs_insert_empty_items
- 1.41% __tracepoint_kvm_entry
0.92% __tracepoint_kvm_apic_accept_irq
- 6.41% relocate_data_extent
- 7.52% relocate_file_extent_cluster
- 2.35% __do_page_cache_readahead
- 1.70% read_pages.isra.0
- 2.03% extent_readpages
1.32% __do_readpage
0.93% __alloc_pages_nodemask
- 1.67% clear_extent_bit
1.92% __clear_extent_bit
- 1.40% lock_extent_bits
1.59% __set_extent_bit
1.19% btrfs_delalloc_reserve_metadata
1.15% __set_page_dirty_nobuffers
- 2.23% prepare_to_merge
- 1.84% btrfs_commit_transaction
- 2.15% btrfs_run_delayed_refs
- __btrfs_run_delayed_refs
1.47% __btrfs_inc_extent_ref.isra.0
+ 38.75% 0.47% [kernel] [k] __btrfs_run_delayed_refs
+ 32.98% 0.21% [kernel] [k] relocate_file_extent_cluster
+ 29.88% 0.00% [kernel] [k] relocate_data_extent
+ 21.63% 0.00% libc-2.31.so [.] __GI___ioctl
+ 21.46% 0.00% [kernel] [k] entry_SYSCALL_64_after_hwframe
+ 21.46% 0.00% [kernel] [k] do_syscall_64
+ 21.46% 0.00% [kernel] [k] __x64_sys_ioctl
+ 21.46% 0.00% [kernel] [k] ksys_ioctl
+ 21.46% 0.00% [kernel] [k] rpc_info_open
+ 21.46% 0.00% [kernel] [k] btrfs_rm_device
For a higher level overview, try: perf top --sort comm,dso
And also running perf doesn't improve performance, which it does
rather remarkably for send/receive.
--
Chris Murphy
next prev parent reply other threads:[~2020-05-06 5:48 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-04 7:43 5.6, slow send/receive, < 1MB/s Chris Murphy
2020-05-04 7:47 ` Nikolay Borisov
2020-05-04 8:03 ` Chris Murphy
2020-05-04 8:12 ` Chris Murphy
2020-05-04 8:13 ` Nikolay Borisov
2020-05-04 8:25 ` Chris Murphy
2020-05-04 8:43 ` Chris Murphy
2020-05-06 5:48 ` Chris Murphy [this message]
2020-05-06 21:04 ` Chris Murphy
2020-05-06 21:07 ` Chris Murphy
2020-05-23 16:11 ` Jorge Bastos
2020-05-23 16:42 ` A L
2020-06-06 16:16 ` Jorge Bastos
2020-06-06 16:24 ` Jorge Bastos
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAJCQCtQu4ffJYuOUWkhV_wR7L0ya7mTyt0tuLqbko-O8S+1fmg@mail.gmail.com \
--to=lists@colorremedies.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=nborisov@suse.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).