Writes to DB disks

[shelton1024]

Hello
Do any of you guys know how much data is written to your Database disks each day?

My database size is ~2.1TB and daily write to the DB disks is about six times the DB size according to SMART info.

Output from one of the DB disks.
SMART/Health Information (NVMe Log 0x02)
Critical Warning: 0x00
Temperature: 51 Celsius
Available Spare: 99%
Available Spare Threshold: 10%
Percentage Used: 98%
Data Units Read: 12,301,370,239 [6.29 PB]
Data Units Written: 8,484,848,376 [4.34 PB]
Host Read Commands: 177,226,070,849
Host Write Commands: 150,042,773,680
Controller Busy Time: 269,175
Power Cycles: 23
Power On Hours: 10,074
Unsafe Shutdowns: 15
Media and Data Integrity Errors: 0
Error Information Log Entries: 0
Warning Comp. Temperature Time: 0
Critical Comp. Temperature Time: 0

Error Information (NVMe Log 0x01, max 64 entries)
No Errors Logged

Thanks

 

[marclant]

I can't tell you if that's high or not, that's not something we normally look at. But I will give you a bit more information on what constitutes a write on disk when looking at it from the Operating System. Anytime data is modified on the disk, that's a write. So when a row is inserted, updated or deleted from the database, that's write operation to disk. If the data is only read, but not modified, then it's a read.

So when the OS reports X number of writes, that corresponds inserts, updates and deletes in the database.

• Expiration does updates and deletes
• Backups do inserts and deletes (the deletes because of older version dropping off)
• Reorgs do a series of inserts and deletes to move data around
• With sequential pools, reclamation, migration also do inserts, updates and deletes
• With sequential pools, if doing a copy to tape, that's additional inserts and updates
• With container pools, each client backup does a lot of insert for each new extents, and updates for existing extents being referenced
• With container pools, protect and replication will do inserts and updates
• With container pools, automatic move containers will do inserts, updates and deletes.

As you can see, there's a whole lot of database activity that can cause writes to disk. And I just listed the common ones, but any process running will cause database records to be inserted, updated and/or deleted. So it adds up.

 

[shelton1024]

Thank you for the reply @marclant

What raises my concern is that the server in question seems to have abnormally high writes compared to the server that was in use before this one.
Looking at a ssd that's from the now retired server I see that it had ~0.55pb written over a 6 year period which equals to ~0.09pb per year where the disks on new server which has been in use for a year now have 4.34pb written on them.

Granted the setup is not exactly the same between the old and new server (rhel6 vs rhel7 for example) the servers do play the same role in our environment so that I did not expect to see this spike in writes.

The new server does have nvme ssd disks which are substantially faster than the sata ssd disks in the old server, could it be that the old disks were a bottleneck and we would have seen this amount of writes on the old server if the disks would have been faster? Or do you suspect there is something wrong with our setup?

Thanks again!

 

[marclant]

I don't suspect something wrong unless you are seeing other symptoms than this. Slower disk would while a bottleneck would not cause less activity though, it would just cause it to run slower.

Any chance newer features were enabled on the new server, like moving to container pool for example. Or started to use the Operation Center.