pcp
[Top] [All Lists]

Re: [pcp] Queston about pcp performance metrics filesys.used

To: Nathan Scott <nathans@xxxxxxxxxx>, William Staten <William_Staten@xxxxxxxxxxxxxxx>
Subject: Re: [pcp] Queston about pcp performance metrics filesys.used
From: Mark Goodwin <mgoodwin@xxxxxxxxxx>
Date: Wed, 10 Feb 2016 13:21:59 +1100
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <1205110488.19797558.1455068828315.JavaMail.zimbra@xxxxxxxxxx>
References: <OF1B9DD255.E1F86B08-ON85257F54.004E7C3F-85257F54.005054A1@xxxxxxxxxxxxxxx> <136347470.19757537.1455053473332.JavaMail.zimbra@xxxxxxxxxx> <OF46C1D315.1830CBD6-ON85257F55.0003D3CE-85257F55.00040FD8@xxxxxxxxxxxxxxx> <1205110488.19797558.1455068828315.JavaMail.zimbra@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.0
On 02/10/2016 12:47 PM, Nathan Scott wrote:


----- Original Message -----
Nathan,

Thanks I will look into this. Any reason why this is done from the disk.dm.
metrics but not the filesys. metrics?


Good question  :)  No real reason there were not separated out AFAICT - Mark,
thoughts?  Memories from way back then?


disk.dm metrics were added relatively recently. The dm-X names are not 
persistent
across reboots and thus generally unsuitable for use as external instance names.
By contrast, the mapped names are persistent, and far more useful for storage
performance analysis. Who hasn't cursed sysstat's device naming? (who hasn't
cursed sysstat for that matter ;)

See the following RH BZ for context:
  BZ 1109539 hinv.map.lvname instance IDs are not persistent

So IMO it would make sense to change the external instance naming for the 
filesys
metrics to map the dm name to the lv name for mounted device-mapper devices. Non
dm block device naming would remain the same. The mapping is readily available,
see map_persistent_dm_name() in src/pmdas/linux/proc_partitions.c

Regards
-- Mark

<Prev in Thread] Current Thread [Next in Thread>