Skip to Content
0

Heavy disk usage in SAP Solution manager installed on ASE database

May 25, 2017 at 04:48 PM

54

avatar image

Hello!

I'm having a problem with the fact that the Solman server continuously heavily loads storage: average disk i/o rate stay around 25 MB / s, increasing when backups are created up to 75-100 MB / s with a peak up to 200+ MB / s At this time, other servers with the ASE database when creating backups have a short-term increase in the exchange rate with a disk of 20 MB / s

Can you please tell me what could be the reasons for the constant load on the database on one particular server?

Disk usage chart:

solman-disk-usage.png

Thank you!

Sergey Taranenko

10 |10000 characters needed characters left characters exceeded
* Please Login or Register to Answer, Follow or Comment.

2 Answers

Best Answer
Mark A Parsons May 25, 2017 at 06:23 PM
0

NOTE: While I work with SAP/Sybase ASE, I do not work with SAP applications (eg, SOLMAN), so fwiw ...

NOTE: I'm assuming this disk usage is measured at the OS level (ie, outside of ASE)

A few things to consider ...

  • how much of that disk usage is writes? reads?
  • if the ASE databases devices are sitting on a filesystem, make sure journaling is disabled on the filesystem [ASE does not need FS journaling; FS journaling generates excessive write actitity]
  • what kind of activity is being performed in ASE? are you seeing any queries performing a lot of disk IOs and if so are they reads or writes? [heavy reads could indicate poor query plans constantly reading large tables from disk - tables that are too big to fit into ASE memory; heavy writes may be normal for a lot of insert/delete/update activity]
  • what kind of disk activity (if any) is being performed against the different ASE database devices, and what databases are said devices related to? [heavy write activity on tempdb devices, especially for short-lived #temp tables and queries, could indicate a need to disable the housekeeper so as to reduce the volume of needless writes to disk]
  • does MDA monitoring show comparable volumes of activity (inserts/deletes/updates/tempdb-activity) between the different servers in question? [eg, a more active dataserver will likely show more disk activity]
  • when comparing dump activity between different servers you need to take into consideration the dump commands (# of stripes, (non)use of compression, size of databases, amount of used space in databases, other dump configs/attributes) [ eg, more stripes means more sybmultbuf processes reading/writing in parallel which could show up as higher IO rates; no compression means more bytes written to dump files]
  • any chance this particular server has both ASE database devices (reads) and the dump files (writes) sitting on the same raw devices, while other servers have the database devices (reads) and dump files (writes) on different raw devices?
  • if the dump commands and database sizes are comparable on the different servers, any chance this server is completing it's dumps in a more timely manner than the other servers? [if this server has a faster disk subsystem then 'faster' dumps would equate to higher IO rates/throughput, while slower dumps would tend to indicate lower IO rates/throughput; net result being that by itself higher IO rates aren't necessarily a bad thing]
Share
10 |10000 characters needed characters left characters exceeded
Sergey Taranenko Jun 06, 2017 at 09:27 AM
0

Hello!
Thank you for advice.

Disk usage graph taken from VMware. At OS level i saw mostrly READ usage.

Currently this problem solved by updating database to latest patch and changing housekeeper GC from agressive to lazy mode.
After updating the database was heavily loaded for 2 days, after the load fell to 5-15 MB/s.

Share
10 |10000 characters needed characters left characters exceeded