Primera upgrade to 4.2.0.59
Posted: Fri Jul 24, 2020 10:56 am
So I wanted to put this out there in case others haven't made the jump yet. As you know us customers can upgrade to the latest code as soon as we desire because of the simplified upgrade process. At our organization we always try to be a little cautious but in the end we are more willing to take a newer upgrade when we know there are enhancements that we need. For 3PAR we were early adopters of the 3.3.1 code because we needed VVs larger than 16TB.
So we had a few "minor" issues with Primera since we had them running in October. We had cages that would alarm because it was "too cold". Right? We had an issue where garbage collection would stop working so when we had a remote 3par replicating to the Primera every 5 minutes, the GC process could't clean up our snapshot space. So our CPGs were running out of room as we had consumed upwards of 700TB of snap space.
The final issue we had was our write performance was not as good as our 3PARs. Reads on the Primera are phenomenal, ~.25ms on average but our write response times were averaging 2ms. We run our primeras in a peer persistence setup so low write latency is the key to making that work.
4.2.0.59 code promised to fix many of these issues and since the code had been out for a few weeks we thought we'd get it installed. So we upgraded the Primera at our secondary DC first. The upgrade...other than taking 2 hrs. vs the 1hr that is advertised went quite smooth. After the upgrade we noticed a greatly improved write performance for the local VVs that were not in a PP relationship. This in itself had us quite excited to get the secondary array upgraded the next evening.
On to the second array I had the initial attempt fail in the pre-checks because of a somewhat cryptic error that I eventually figured out was a warning about a host not being zoned to all the controllers. My bad. Once that was fixed i restarted the upgrade process and it continued past that warning. I had reached out to support before i figured out the issue and about 20 minutes later I was invited to a myroom session. Since I figured it out, i was planning on just letting support know that i didn't need them anymore.
This is where things got a little scary. His response was "oh, I thought they pulled back all the 4.2 upgrades because of all the issues." Right? He assured me that he saw many successful upgrades but there were also many customers that ran into issues.
So the good news on the upgrade is that my second upgrade finished in just under 2 hrs. without issue. The write performance has been dramatically increased and now I can truly say that the Primera is faster than the 3par. We have had zero issues and things are working great! I was assured that new code would be released soon to fix the upgrade issues.
The bad news is some other customers were not so lucky. Some even had failures. So obviously hold off on the upgrade until they release new code and I know next time I do an upgrade I will ask HPE to make sure I get an all clear.
So we had a few "minor" issues with Primera since we had them running in October. We had cages that would alarm because it was "too cold". Right? We had an issue where garbage collection would stop working so when we had a remote 3par replicating to the Primera every 5 minutes, the GC process could't clean up our snapshot space. So our CPGs were running out of room as we had consumed upwards of 700TB of snap space.
The final issue we had was our write performance was not as good as our 3PARs. Reads on the Primera are phenomenal, ~.25ms on average but our write response times were averaging 2ms. We run our primeras in a peer persistence setup so low write latency is the key to making that work.
4.2.0.59 code promised to fix many of these issues and since the code had been out for a few weeks we thought we'd get it installed. So we upgraded the Primera at our secondary DC first. The upgrade...other than taking 2 hrs. vs the 1hr that is advertised went quite smooth. After the upgrade we noticed a greatly improved write performance for the local VVs that were not in a PP relationship. This in itself had us quite excited to get the secondary array upgraded the next evening.
On to the second array I had the initial attempt fail in the pre-checks because of a somewhat cryptic error that I eventually figured out was a warning about a host not being zoned to all the controllers. My bad. Once that was fixed i restarted the upgrade process and it continued past that warning. I had reached out to support before i figured out the issue and about 20 minutes later I was invited to a myroom session. Since I figured it out, i was planning on just letting support know that i didn't need them anymore.
This is where things got a little scary. His response was "oh, I thought they pulled back all the 4.2 upgrades because of all the issues." Right? He assured me that he saw many successful upgrades but there were also many customers that ran into issues.
So the good news on the upgrade is that my second upgrade finished in just under 2 hrs. without issue. The write performance has been dramatically increased and now I can truly say that the Primera is faster than the 3par. We have had zero issues and things are working great! I was assured that new code would be released soon to fix the upgrade issues.
The bad news is some other customers were not so lucky. Some even had failures. So obviously hold off on the upgrade until they release new code and I know next time I do an upgrade I will ask HPE to make sure I get an all clear.