Hi!

I’m getting

```
Error summary: IllegalStateException: ARPACK returns non-zero info = 1 Maximum number of iterations taken. (Refer ARPACK user guide for details)
```

after calling `hwe_normalized_pca(mt.GT, k=2)`

and I have trouble pinpointing the cause of the problem. What’s worse, when I’m calling this function several times in a row on the same input, sometimes it errors out and sometimes it returns without an error. Is that even possible with a proper installation? I have now made myself a jupyter notebook with the following single cell just to prove to myself that I’m not going mad:

```
hl.init(quiet=True)
hl.set_global_seed(42)
mt = hl.read_matrix_table(my_path)
for i in range(12):
try:
hl.hwe_normalized_pca(
mt.GT,
k = 2
)
print(f"{i} success")
except:
print(f"{i} error")
```

and it prints

```
0 success
1 success
2 success
3 success
4 error
5 success
6 success
7 success
8 success
9 error
10 success
11 success
```

where `mt`

is a product of a several filtering steps performed on a genuine vcf.

Also, when I do

```
hl.pca(
hl.int(hl.is_defined(mt.GT)),
k = 2
)
```

on my `mt`

, I always immediately get another kind of error:

`Error summary: HailException: Cannot create RowMatrix: filtered entry at row 8111 and col 9`

,

but I don’t know if it is related to the problem with `hwe_normalizd_pca`

.

Any suggestions would be appreciated:

- What qualities of any
`mt`

could in general cause the`hwe_normalized_pca`

to not converge? - Should I suspect that there’s something wrong with my installation? I’m running hail locally, on a laptop.
- Shape of
`mt`

is (18205, 39), its only non-key field is GT, and the col-keys are generic strings, so I’d be happy to share it.