From: | Jeevan Chalke <jeevan(dot)chalke(at)enterprisedb(dot)com> |
---|---|
To: | vignesh C <vignesh21(at)gmail(dot)com> |
Cc: | Ibrar Ahmed <ibrar(dot)ahmad(at)gmail(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>, Anastasia Lubennikova <a(dot)lubennikova(at)postgrespro(dot)ru>, Stephen Frost <sfrost(at)snowman(dot)net>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | Re: block-level incremental backup |
Date: | 2019-09-09 11:21:34 |
Message-ID: | CAM2+6=VF3VsR4LJ451qYYaE6MqfY73GohWCaBWOeyvsVb_UXHg@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Tue, Aug 27, 2019 at 4:46 PM vignesh C <vignesh21(at)gmail(dot)com> wrote:
> Few comments:
> Comment:
> + buf = (char *) malloc(statbuf->st_size);
> + if (buf == NULL)
> + ereport(ERROR,
> + (errcode(ERRCODE_OUT_OF_MEMORY),
> + errmsg("out of memory")));
> +
> + if ((cnt = fread(buf, 1, statbuf->st_size, fp)) > 0)
> + {
> + Bitmapset *mod_blocks = NULL;
> + int nmodblocks = 0;
> +
> + if (cnt % BLCKSZ != 0)
> + {
>
> We can use same size as full page size.
> After pg start backup full page write will be enabled.
> We can use the same file size to maintain data consistency.
>
Can you please explain which size?
The aim here is to read entire file in-memory and thus used
statbuf->st_size.
Comment:
> Should we check if it is same timeline as the system's timeline.
>
At the time of taking the incremental backup, we can't check that.
However, while combining, I made sure that the timeline is the same for all
backups.
>
> Comment:
>
> Should we support compression formats supported by pg_basebackup.
> This can be an enhancement after the functionality is completed.
>
For the incremental backup, it just works out of the box.
For combining backup, as discussed up-thread, the user has to
uncompress first, combine them, compress if required.
> Comment:
> We should provide some mechanism to validate the backup. To identify
> if some backup is corrupt or some file is missing(deleted) in a
> backup.
>
Maybe, but not for the first version.
> Comment:
> +/*
> + * When to send the whole file, % blocks modified (90%)
> + */
> +#define WHOLE_FILE_THRESHOLD 0.9
> +
> This can be user configured value.
> This can be an enhancement after the functionality is completed.
>
Yes.
> Comment:
> We can add a readme file with all the details regarding incremental
> backup and combine backup.
>
Will have a look.
>
> Regards,
> Vignesh
> EnterpriseDB: http://www.enterprisedb.com
>
Thanks
--
Jeevan Chalke
Technical Architect, Product Development
EnterpriseDB Corporation
The Enterprise PostgreSQL Company
From | Date | Subject | |
---|---|---|---|
Next Message | Jeevan Chalke | 2019-09-09 11:30:33 | Re: block-level incremental backup |
Previous Message | Jeevan Chalke | 2019-09-09 11:08:15 | Re: block-level incremental backup |