Disclaimer, the command is "fast", but it takes a long time to reclaim the space(could take weeks to reclaim everything). Run unmap command from ESXi(esxcli storage vmfs unmap something somethinf if my menory serves me right). With UNMAP, the correct way is thin on thin.ģ.
Without UNMAP in Vmware (require VMFS6 and ESXi 6.5 or later), the correct way is thin (both dedupe and compression are thin volumes) on 3PAR and Eager Zero Thick in Vmware. With increasingly larger SSDs this is a serious issue. you just need a single disk failure and an unrecoverable read error on any of the other drives in that RAID set to be screwed. Remember that you don't need a dual disk failure to have an issue with RAID5. I don't think you find a single vendor in the industry recommending anything with less than double parity today. What have I forgotten ? Any other tips how to maximize compression and dedup on 3Par ? Big thanks for your time and advice in advance !ġ. This way, thin allocation of disk space will be shifted from virtual machine/VMware level to 3Par storage level but we are fine with that. Overwriting unused space within virtual machines with 0s will effectively inflate their thin provisioned disks on VMware level, so we have to have enough space available on datastore levels. let it sit overnight or two days, so 3Par ASIC can do its job and detect 0s reclaim unused space on VMware datastore level (esxcli storage vmfs unmap -reclaim-unit=200 -volume-label=XXXXX) Several OS-dependent tools available to do this. overwrite unused space within virtual machines with zeros. delete unnecessary data within virtual machines (yay, sounds easy). What would potentially be the best way how to deal with the situation ? Over the time, we accumulated lot of ballast. Model 20850, 32x 4TB SSD, OS 3.2.2MU6, single CPG (Raid5 due to space efficiency, we didn't need anything else performance is very fine - we are hardly getting below 30.000 IOPS all day long, quite usually above 150.000 IOPS and 3GB/s and almost always with sub-1ms latencies). Some 3Par VVs are thin, majority is dedup. 100% virtualized environment, thin provisioning on virtual machines. We operate VMware 5.5 and 6.0 environments with several Win2008 and CentOS/linux legacy systems so TRIM functionality is a problem for us. Guys, help a noob with the best space reclamation please.
0 kommentar(er)
