4
4
4ebyanka2019-12-17 11:08:42
Data transfer
4ebyanka, 2019-12-17 11:08:42

Why such a slow sequential read through FiberChannel?

Tell me, maybe someone works with SAN Lenovo (de4000h 2u24 sff) and knows them well. There is a Lenovo server with ESXi 6.7 and a SAN with 24 disks of 1.8 TB each (sas 12g 10k hdd) in 10 RAID is connected to it directly via FC. At first I didn’t understand why Ubuntu reads from it so slowly, it turned out that I couldn’t read more than 160 MB per second at all (I check on a 10GB file). Nearby is the same server with the same disks, only inside the server (8 disks in 10 raids) and there reading is already 900mb per second with the same file and with the same system. In the SAN settings, it is indicated everywhere that this is for VMware and the volume is also created taking into account VMware vmfs, there is no ssd cache, the block size is 64K

Answer the question

In order to leave comments, you need to log in

1 answer(s)
N
noute, 2020-01-10
@noute

Difficult question - it is necessary to analyze
1. SHD performance
2. SAN network
performance 3. Server
performance 4. OS performance.
5. Multipass settings and log analysis
Block size 64 K where - SHD, Datastore, File?
Look, maybe the datastore is clogged, so this is the situation. The file system may also be corrupted.
Is there only one server on these disks or is there something that is spinning in the neighborhood?
In what provisions are datastores created?
Thin provising - if there is a large increase in the server, the system allocates disk space and until it fills it up, there will be brakes, in other words, with this provision, there may be a problem with growth.
In Thin provising zeried - there is no such problem, but there is a bicycle there.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question