Re: pg_stat_progress_basebackup - progress reporting for pg_basebackup, in the server side

From: Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>
To: masao(dot)fujii(at)oss(dot)nttdata(dot)com
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: pg_stat_progress_basebackup - progress reporting for pg_basebackup, in the server side
Date: 2020-01-30 03:58:41
Message-ID: 20200130.125841.2006218843084031993.horikyota.ntt@gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

At Wed, 29 Jan 2020 23:16:08 +0900, Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> wrote in
> Hi,
>
> pg_basebackup reports the backup progress if --progress option is
> specified,
> and we can monitor it in the client side. I think that it's useful if
> we can
> monitor the progress information also in the server side because, for
> example,
> we can easily check that by using SQL. So I'd like to propose
> pg_stat_progress_basebackup view that allows us to monitor the
> progress
> of pg_basebackup, in the server side. Thought?
>
> POC patch is attached.

| postgres=# \d pg_stat_progress_basebackup
| View "pg_catalog.pg_stat_progress_basebackup"
| Column | Type | Collation | Nullable | Default
| ---------------------+---------+-----------+----------+---------
| pid | integer | | |
| phase | text | | |
| backup_total | bigint | | |
| backup_streamed | bigint | | |
| tablespace_total | bigint | | |
| tablespace_streamed | bigint | | |

I think the view needs client identity such like host/port pair
besides pid (host/port pair fails identify client in the case of
unix-sockets.). Also elapsed time from session start might be
useful. pg_stat_progress_acuum has datid, datname and relid.

+ if (backup_total > 0 && backup_streamed > backup_total)
+ {
+ backup_total = backup_streamed;

Do we need the condition "backup_total > 0"?

+ int tblspc_streamed = 0;
+
+ pgstat_progress_update_param(PROGRESS_BASEBACKUP_PHASE,
+ PROGRESS_BASEBACKUP_PHASE_STREAM_BACKUP);

This starts "streaming backup" phase with backup_total = 0. Coudln't
we move to that phase after setting backup total and tablespace total?
That is, just after calling SendBackupHeader().

+ WHEN 3 THEN 'stopping backup'::text

I'm not sure, but the "stop" seems suggesting the backup is terminated
before completion. If it is following the name of the function
pg_stop_backup, I think the name is suggesting to stop "the state for
performing backup", not a backup.

In the first place, the progress is about "backup" so it seems strange
that we have another phase after the "stopping backup" phase. It
might be better that it is "finishing file transfer" or such.

"initializing"
-> "starting file transfer"
-> "transferring files"
-> "finishing file transfer"
-> "transaferring WAL"

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Kyotaro Horiguchi 2020-01-30 04:01:39 Re: pause recovery if pitr target not reached
Previous Message Amit Kapila 2020-01-30 03:34:48 Re: closesocket behavior in different platforms